tern coverage - 95.63%

Files
Source

#' Count patients with toxicity grades that have worsened from baseline by highest grade post-baseline
#'
#' @description `r lifecycle::badge("stable")`
#'
#' The analyze function [count_abnormal_lab_worsen_by_baseline()] creates a layout element to count patients with
#' analysis toxicity grades which have worsened from baseline, categorized by highest (worst) grade post-baseline.
#'
#' This function analyzes primary analysis variable `var` which indicates analysis toxicity grades. Additional
#' analysis variables that can be supplied as a list via the `variables` parameter are `id` (defaults to `USUBJID`),
#' a variable to indicate unique subject identifiers, `baseline_var` (defaults to `BTOXGR`), a variable to indicate
#' baseline toxicity grades, and `direction_var` (defaults to `GRADDIR`), a variable to indicate toxicity grade
#' directions of interest to include (e.g. `"H"` (high), `"L"` (low), or `"B"` (both)).
#'
#' For the direction(s) specified in `direction_var`, patient counts by worst grade for patients who have
#' worsened from baseline are calculated as follows:
#'   * `1` to `4`: The number of patients who have worsened from their baseline grades with worst
#'     grades 1-4, respectively.
#'   * `Any`: The total number of patients who have worsened from their baseline grades.
#'
#' Fractions are calculated by dividing the above counts by the number of patients who's analysis toxicity grades
#' have worsened from baseline toxicity grades during treatment.
#'
#' Prior to using this function in your table layout you must use [rtables::split_rows_by()] to create a row
#' split on variable `direction_var`.
#'
#' @inheritParams argument_convention
#' @param variables (named `list` of `string`)\cr list of additional analysis variables including:
#'   * `id` (`string`)\cr subject variable name.
#'   * `baseline_var` (`string`)\cr name of the data column containing baseline toxicity variable.
#'   * `direction_var` (`string`)\cr see `direction_var` for more details.
#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("abnormal_by_worst_grade_worsen")`
#'   to see all available statistics.
#'
#' @seealso Relevant helper functions [h_adlb_worsen()] and [h_worsen_counter()] which are used within
#' [s_count_abnormal_lab_worsen_by_baseline()] to process input data.
#'
#' @name abnormal_by_worst_grade_worsen
#' @order 1
NULL

#' Helper function to prepare ADLB with worst labs
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Helper function to prepare a `df` for generate the patient count shift table.
#'
#' @param adlb (`data.frame`)\cr ADLB data frame.
#' @param worst_flag_low (named `vector`)\cr worst low post-baseline lab grade flag variable. See how this is
#'   implemented in the following examples.
#' @param worst_flag_high (named `vector`)\cr worst high post-baseline lab grade flag variable. See how this is
#'   implemented in the following examples.
#' @param direction_var (`string`)\cr name of the direction variable specifying the direction of the shift table of
#'   interest. Only lab records flagged by `L`, `H` or `B` are included in the shift table.
#'   * `L`: low direction only
#'   * `H`: high direction only
#'   * `B`: both low and high directions
#'
#' @return `h_adlb_worsen()` returns the `adlb` `data.frame` containing only the
#'   worst labs specified according to `worst_flag_low` or `worst_flag_high` for the
#'   direction specified according to `direction_var`. For instance, for a lab that is
#'   needed for the low direction only, only records flagged by `worst_flag_low` are
#'   selected. For a lab that is needed for both low and high directions, the worst
#'   low records are selected for the low direction, and the worst high record are selected
#'   for the high direction.
#'
#' @seealso [abnormal_by_worst_grade_worsen]
#'
#' @examples
#' library(dplyr)
#'
#' # The direction variable, GRADDR, is based on metadata
#' adlb <- tern_ex_adlb %>%
#'   mutate(
#'     GRADDR = case_when(
#'       PARAMCD == "ALT" ~ "B",
#'       PARAMCD == "CRP" ~ "L",
#'       PARAMCD == "IGA" ~ "H"
#'     )
#'   ) %>%
#'   filter(SAFFL == "Y" & ONTRTFL == "Y" & GRADDR != "")
#'
#' df <- h_adlb_worsen(
#'   adlb,
#'   worst_flag_low = c("WGRLOFL" = "Y"),
#'   worst_flag_high = c("WGRHIFL" = "Y"),
#'   direction_var = "GRADDR"
#' )
#'
#' @export
h_adlb_worsen <- function(adlb,
                          worst_flag_low = NULL,
                          worst_flag_high = NULL,
                          direction_var) {
  checkmate::assert_string(direction_var)
  checkmate::assert_subset(as.character(unique(adlb[[direction_var]])), c("B", "L", "H"))
  assert_df_with_variables(adlb, list("Col" = direction_var))

  if (any(unique(adlb[[direction_var]]) == "H")) {
    assert_df_with_variables(adlb, list("High" = names(worst_flag_high)))
  }

  if (any(unique(adlb[[direction_var]]) == "L")) {
    assert_df_with_variables(adlb, list("Low" = names(worst_flag_low)))
  }

  if (any(unique(adlb[[direction_var]]) == "B")) {
    assert_df_with_variables(
      adlb,
      list(
        "Low" = names(worst_flag_low),
        "High" = names(worst_flag_high)
      )
    )
  }

  # extract patients with worst post-baseline lab, either low or high or both
  worst_flag <- c(worst_flag_low, worst_flag_high)
  col_names <- names(worst_flag)
  filter_values <- worst_flag
  temp <- Map(
    function(x, y) which(adlb[[x]] == y),
    col_names,
    filter_values
  )
  position_satisfy_filters <- Reduce(union, temp)

  # select variables of interest
  adlb_f <- adlb[position_satisfy_filters, ]

  # generate subsets for different directionality
  adlb_f_h <- adlb_f[which(adlb_f[[direction_var]] == "H"), ]
  adlb_f_l <- adlb_f[which(adlb_f[[direction_var]] == "L"), ]
  adlb_f_b <- adlb_f[which(adlb_f[[direction_var]] == "B"), ]

  # for labs requiring both high and low, data is duplicated and will be stacked on top of each other
  adlb_f_b_h <- adlb_f_b
  adlb_f_b_l <- adlb_f_b

  # extract data with worst lab
  if (!is.null(worst_flag_high) && !is.null(worst_flag_low)) {
    # change H to High, L to Low
    adlb_f_h[[direction_var]] <- rep("High", nrow(adlb_f_h))
    adlb_f_l[[direction_var]] <- rep("Low", nrow(adlb_f_l))

    # change, B to High and Low
    adlb_f_b_h[[direction_var]] <- rep("High", nrow(adlb_f_b_h))
    adlb_f_b_l[[direction_var]] <- rep("Low", nrow(adlb_f_b_l))

    adlb_out_h <- adlb_f_h[which(adlb_f_h[[names(worst_flag_high)]] == worst_flag_high), ]
    adlb_out_b_h <- adlb_f_b_h[which(adlb_f_b_h[[names(worst_flag_high)]] == worst_flag_high), ]
    adlb_out_l <- adlb_f_l[which(adlb_f_l[[names(worst_flag_low)]] == worst_flag_low), ]
    adlb_out_b_l <- adlb_f_b_l[which(adlb_f_b_l[[names(worst_flag_low)]] == worst_flag_low), ]

    out <- rbind(adlb_out_h, adlb_out_b_h, adlb_out_l, adlb_out_b_l)
  } else if (!is.null(worst_flag_high)) {
    adlb_f_h[[direction_var]] <- rep("High", nrow(adlb_f_h))
    adlb_f_b_h[[direction_var]] <- rep("High", nrow(adlb_f_b_h))

    adlb_out_h <- adlb_f_h[which(adlb_f_h[[names(worst_flag_high)]] == worst_flag_high), ]
    adlb_out_b_h <- adlb_f_b_h[which(adlb_f_b_h[[names(worst_flag_high)]] == worst_flag_high), ]

    out <- rbind(adlb_out_h, adlb_out_b_h)
  } else if (!is.null(worst_flag_low)) {
    adlb_f_l[[direction_var]] <- rep("Low", nrow(adlb_f_l))
    adlb_f_b_l[[direction_var]] <- rep("Low", nrow(adlb_f_b_l))

    adlb_out_l <- adlb_f_l[which(adlb_f_l[[names(worst_flag_low)]] == worst_flag_low), ]
    adlb_out_b_l <- adlb_f_b_l[which(adlb_f_b_l[[names(worst_flag_low)]] == worst_flag_low), ]

    out <- rbind(adlb_out_l, adlb_out_b_l)
  }

  # label
  formatters::var_labels(out) <- formatters::var_labels(adlb_f, fill = FALSE)
  # NA
  out
}

#' Helper function to analyze patients for `s_count_abnormal_lab_worsen_by_baseline()`
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Helper function to count the number of patients and the fraction of patients according to
#' highest post-baseline lab grade variable `.var`, baseline lab grade variable `baseline_var`,
#' and the direction of interest specified in `direction_var`.
#'
#' @inheritParams argument_convention
#' @inheritParams h_adlb_worsen
#' @param baseline_var (`string`)\cr name of the baseline lab grade variable.
#'
#' @return The counts and fraction of patients
#'   whose worst post-baseline lab grades are worse than their baseline grades, for
#'   post-baseline worst grades "1", "2", "3", "4" and "Any".
#'
#' @seealso [abnormal_by_worst_grade_worsen]
#'
#' @examples
#' library(dplyr)
#'
#' # The direction variable, GRADDR, is based on metadata
#' adlb <- tern_ex_adlb %>%
#'   mutate(
#'     GRADDR = case_when(
#'       PARAMCD == "ALT" ~ "B",
#'       PARAMCD == "CRP" ~ "L",
#'       PARAMCD == "IGA" ~ "H"
#'     )
#'   ) %>%
#'   filter(SAFFL == "Y" & ONTRTFL == "Y" & GRADDR != "")
#'
#' df <- h_adlb_worsen(
#'   adlb,
#'   worst_flag_low = c("WGRLOFL" = "Y"),
#'   worst_flag_high = c("WGRHIFL" = "Y"),
#'   direction_var = "GRADDR"
#' )
#'
#' # `h_worsen_counter`
#' h_worsen_counter(
#'   df %>% filter(PARAMCD == "CRP" & GRADDR == "Low"),
#'   id = "USUBJID",
#'   .var = "ATOXGR",
#'   baseline_var = "BTOXGR",
#'   direction_var = "GRADDR"
#' )
#'
#' @export
h_worsen_counter <- function(df, id, .var, baseline_var, direction_var) {
  checkmate::assert_string(id)
  checkmate::assert_string(.var)
  checkmate::assert_string(baseline_var)
  checkmate::assert_scalar(unique(df[[direction_var]]))
  checkmate::assert_subset(unique(df[[direction_var]]), c("High", "Low"))
  assert_df_with_variables(df, list(val = c(id, .var, baseline_var, direction_var)))

  # remove post-baseline missing
  df <- df[df[[.var]] != "<Missing>", ]

  # obtain directionality
  direction <- unique(df[[direction_var]])

  if (direction == "Low") {
    grade <- -1:-4
    worst_grade <- -4
  } else if (direction == "High") {
    grade <- 1:4
    worst_grade <- 4
  }

  if (nrow(df) > 0) {
    by_grade <- lapply(grade, function(i) {
      # filter baseline values that is less than i or <Missing>
      df_temp <- df[df[[baseline_var]] %in% c((i + sign(i) * -1):(-1 * worst_grade), "<Missing>"), ]
      # num: number of patients with post-baseline worst lab equal to i
      num <- length(unique(df_temp[df_temp[[.var]] %in% i, id, drop = TRUE]))
      # denom: number of patients with baseline values less than i or <missing> and post-baseline in the same direction
      denom <- length(unique(df_temp[[id]]))
      rm(df_temp)
      c(num = num, denom = denom)
    })
  } else {
    by_grade <- lapply(1, function(i) {
      c(num = 0, denom = 0)
    })
  }

  names(by_grade) <- as.character(seq_along(by_grade))

  # baseline grade less 4 or missing
  df_temp <- df[!df[[baseline_var]] %in% worst_grade, ]

  # denom: number of patients with baseline values less than 4 or <missing> and post-baseline in the same direction
  denom <- length(unique(df_temp[, id, drop = TRUE]))

  # condition 1: missing baseline and in the direction of abnormality
  con1 <- which(df_temp[[baseline_var]] == "<Missing>" & df_temp[[.var]] %in% grade)
  df_temp_nm <- df_temp[which(df_temp[[baseline_var]] != "<Missing>" & df_temp[[.var]] %in% grade), ]

  # condition 2: if post-baseline values are present then post-baseline values must be worse than baseline
  if (direction == "Low") {
    con2 <- which(as.numeric(as.character(df_temp_nm[[.var]])) < as.numeric(as.character(df_temp_nm[[baseline_var]])))
  } else {
    con2 <- which(as.numeric(as.character(df_temp_nm[[.var]])) > as.numeric(as.character(df_temp_nm[[baseline_var]])))
  }

  # number of patients satisfy either conditions 1 or 2
  num <- length(unique(df_temp[union(con1, con2), id, drop = TRUE]))

  list(fraction = c(by_grade, list("Any" = c(num = num, denom = denom))))
}

#' @describeIn abnormal_by_worst_grade_worsen Statistics function for patients whose worst post-baseline
#'   lab grades are worse than their baseline grades.
#'
#' @return
#' * `s_count_abnormal_lab_worsen_by_baseline()` returns the counts and fraction of patients whose worst
#'   post-baseline lab grades are worse than their baseline grades, for post-baseline worst grades
#'   "1", "2", "3", "4" and "Any".
#'
#' @keywords internal
s_count_abnormal_lab_worsen_by_baseline <- function(df, # nolint
                                                    .var = "ATOXGR",
                                                    variables = list(
                                                      id = "USUBJID",
                                                      baseline_var = "BTOXGR",
                                                      direction_var = "GRADDR"
                                                    )) {
  checkmate::assert_string(.var)
  checkmate::assert_set_equal(names(variables), c("id", "baseline_var", "direction_var"))
  checkmate::assert_string(variables$id)
  checkmate::assert_string(variables$baseline_var)
  checkmate::assert_string(variables$direction_var)
  assert_df_with_variables(df, c(aval = .var, variables[1:3]))
  assert_list_of_variables(variables)

  h_worsen_counter(df, variables$id, .var, variables$baseline_var, variables$direction_var)
}

#' @describeIn abnormal_by_worst_grade_worsen Formatted analysis function which is used as `afun`
#'   in `count_abnormal_lab_worsen_by_baseline()`.
#'
#' @return
#' * `a_count_abnormal_lab_worsen_by_baseline()` returns the corresponding list with
#'   formatted [rtables::CellValue()].
#'
#' @keywords internal
a_count_abnormal_lab_worsen_by_baseline <- make_afun( # nolint
  s_count_abnormal_lab_worsen_by_baseline,
  .formats = c(fraction = format_fraction),
  .ungroup_stats = "fraction"
)

#' @describeIn abnormal_by_worst_grade_worsen Layout-creating function which can take statistics function
#'   arguments and additional format arguments. This function is a wrapper for [rtables::analyze()].
#'
#' @return
#' * `count_abnormal_lab_worsen_by_baseline()` returns a layout object suitable for passing to further layouting
#'   functions, or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted
#'   rows containing the statistics from `s_count_abnormal_lab_worsen_by_baseline()` to the table layout.
#'
#' @examples
#' library(dplyr)
#'
#' # The direction variable, GRADDR, is based on metadata
#' adlb <- tern_ex_adlb %>%
#'   mutate(
#'     GRADDR = case_when(
#'       PARAMCD == "ALT" ~ "B",
#'       PARAMCD == "CRP" ~ "L",
#'       PARAMCD == "IGA" ~ "H"
#'     )
#'   ) %>%
#'   filter(SAFFL == "Y" & ONTRTFL == "Y" & GRADDR != "")
#'
#' df <- h_adlb_worsen(
#'   adlb,
#'   worst_flag_low = c("WGRLOFL" = "Y"),
#'   worst_flag_high = c("WGRHIFL" = "Y"),
#'   direction_var = "GRADDR"
#' )
#'
#' basic_table() %>%
#'   split_cols_by("ARMCD") %>%
#'   add_colcounts() %>%
#'   split_rows_by("PARAMCD") %>%
#'   split_rows_by("GRADDR") %>%
#'   count_abnormal_lab_worsen_by_baseline(
#'     var = "ATOXGR",
#'     variables = list(
#'       id = "USUBJID",
#'       baseline_var = "BTOXGR",
#'       direction_var = "GRADDR"
#'     )
#'   ) %>%
#'   append_topleft("Direction of Abnormality") %>%
#'   build_table(df = df, alt_counts_df = tern_ex_adsl)
#'
#' @export
#' @order 2
count_abnormal_lab_worsen_by_baseline <- function(lyt, # nolint
                                                  var,
                                                  variables = list(
                                                    id = "USUBJID",
                                                    baseline_var = "BTOXGR",
                                                    direction_var = "GRADDR"
                                                  ),
                                                  na_str = default_na_str(),
                                                  nested = TRUE,
                                                  ...,
                                                  table_names = NULL,
                                                  .stats = NULL,
                                                  .formats = NULL,
                                                  .labels = NULL,
                                                  .indent_mods = NULL) {
  checkmate::assert_string(var)

  extra_args <- list(variables = variables, ...)

  afun <- make_afun(
    a_count_abnormal_lab_worsen_by_baseline,
    .stats = .stats,
    .formats = .formats,
    .labels = .labels,
    .indent_mods = .indent_mods
  )

  lyt <- analyze(
    lyt = lyt,
    vars = var,
    afun = afun,
    na_str = na_str,
    nested = nested,
    extra_args = extra_args,
    show_labels = "hidden"
  )

  lyt
}

#' Helper function for deriving analysis datasets for select laboratory tables
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Helper function that merges ADSL and ADLB datasets so that missing lab test records are inserted in the
#' output dataset. Remember that `na_level` must match the needed pre-processing
#' done with [df_explicit_na()] to have the desired output.
#'
#' @param adsl (`data.frame`)\cr ADSL data frame.
#' @param adlb (`data.frame`)\cr ADLB data frame.
#' @param worst_flag (named `character`)\cr worst post-baseline lab flag variable. See how this is implemented in the
#'   following examples.
#' @param by_visit (`flag`)\cr defaults to `FALSE` to generate worst grade per patient.
#'   If worst grade per patient per visit is specified for `worst_flag`, then
#'   `by_visit` should be `TRUE` to generate worst grade patient per visit.
#' @param no_fillin_visits (named `character`)\cr visits that are not considered for post-baseline worst toxicity
#'   grade. Defaults to `c("SCREENING", "BASELINE")`.
#'
#' @return `df` containing variables shared between `adlb` and `adsl` along with variables `PARAM`, `PARAMCD`,
#'   `ATOXGR`, and `BTOXGR` relevant for analysis. Optionally, `AVISIT` are `AVISITN` are included when
#'   `by_visit = TRUE` and `no_fillin_visits = c("SCREENING", "BASELINE")`.
#'
#' @details In the result data missing records will be created for the following situations:
#'   * Patients who are present in `adsl` but have no lab data in `adlb` (both baseline and post-baseline).
#'   * Patients who do not have any post-baseline lab values.
#'   * Patients without any post-baseline values flagged as the worst.
#'
#' @examples
#' # `h_adsl_adlb_merge_using_worst_flag`
#' adlb_out <- h_adsl_adlb_merge_using_worst_flag(
#'   tern_ex_adsl,
#'   tern_ex_adlb,
#'   worst_flag = c("WGRHIFL" = "Y")
#' )
#'
#' # `h_adsl_adlb_merge_using_worst_flag` by visit example
#' adlb_out_by_visit <- h_adsl_adlb_merge_using_worst_flag(
#'   tern_ex_adsl,
#'   tern_ex_adlb,
#'   worst_flag = c("WGRLOVFL" = "Y"),
#'   by_visit = TRUE
#' )
#'
#' @export
h_adsl_adlb_merge_using_worst_flag <- function(adsl, # nolint
                                               adlb,
                                               worst_flag = c("WGRHIFL" = "Y"),
                                               by_visit = FALSE,
                                               no_fillin_visits = c("SCREENING", "BASELINE")) {
  col_names <- names(worst_flag)
  filter_values <- worst_flag

  temp <- Map(
    function(x, y) which(adlb[[x]] == y),
    col_names,
    filter_values
  )

  position_satisfy_filters <- Reduce(intersect, temp)

  adsl_adlb_common_columns <- intersect(colnames(adsl), colnames(adlb))
  columns_from_adlb <- c("USUBJID", "PARAM", "PARAMCD", "AVISIT", "AVISITN", "ATOXGR", "BTOXGR")

  adlb_f <- adlb[position_satisfy_filters, ] %>%
    dplyr::filter(!.data[["AVISIT"]] %in% no_fillin_visits)
  adlb_f <- adlb_f[, columns_from_adlb]

  avisits_grid <- adlb %>%
    dplyr::filter(!.data[["AVISIT"]] %in% no_fillin_visits) %>%
    dplyr::pull(.data[["AVISIT"]]) %>%
    unique()

  if (by_visit) {
    adsl_lb <- expand.grid(
      USUBJID = unique(adsl$USUBJID),
      AVISIT = avisits_grid,
      PARAMCD = unique(adlb$PARAMCD)
    )

    adsl_lb <- adsl_lb %>%
      dplyr::left_join(unique(adlb[c("AVISIT", "AVISITN")]), by = "AVISIT") %>%
      dplyr::left_join(unique(adlb[c("PARAM", "PARAMCD")]), by = "PARAMCD")

    adsl1 <- adsl[, adsl_adlb_common_columns]
    adsl_lb <- adsl1 %>% merge(adsl_lb, by = "USUBJID")

    by_variables_from_adlb <- c("USUBJID", "AVISIT", "AVISITN", "PARAMCD", "PARAM")

    adlb_btoxgr <- adlb %>%
      dplyr::select(c("USUBJID", "PARAMCD", "BTOXGR")) %>%
      unique() %>%
      dplyr::rename("BTOXGR_MAP" = "BTOXGR")

    adlb_out <- merge(
      adlb_f,
      adsl_lb,
      by = by_variables_from_adlb,
      all = TRUE,
      sort = FALSE
    )
    adlb_out <- adlb_out %>%
      dplyr::left_join(adlb_btoxgr, by = c("USUBJID", "PARAMCD")) %>%
      dplyr::mutate(BTOXGR = .data$BTOXGR_MAP) %>%
      dplyr::select(-"BTOXGR_MAP")

    adlb_var_labels <- c(
      formatters::var_labels(adlb[by_variables_from_adlb]),
      formatters::var_labels(adlb[columns_from_adlb[!columns_from_adlb %in% by_variables_from_adlb]]),
      formatters::var_labels(adsl[adsl_adlb_common_columns[adsl_adlb_common_columns != "USUBJID"]])
    )
  } else {
    adsl_lb <- expand.grid(
      USUBJID = unique(adsl$USUBJID),
      PARAMCD = unique(adlb$PARAMCD)
    )

    adsl_lb <- adsl_lb %>% dplyr::left_join(unique(adlb[c("PARAM", "PARAMCD")]), by = "PARAMCD")

    adsl1 <- adsl[, adsl_adlb_common_columns]
    adsl_lb <- adsl1 %>% merge(adsl_lb, by = "USUBJID")

    by_variables_from_adlb <- c("USUBJID", "PARAMCD", "PARAM")

    adlb_out <- merge(
      adlb_f,
      adsl_lb,
      by = by_variables_from_adlb,
      all = TRUE,
      sort = FALSE
    )

    adlb_var_labels <- c(
      formatters::var_labels(adlb[by_variables_from_adlb]),
      formatters::var_labels(adlb[columns_from_adlb[!columns_from_adlb %in% by_variables_from_adlb]]),
      formatters::var_labels(adsl[adsl_adlb_common_columns[adsl_adlb_common_columns != "USUBJID"]])
    )
  }

  adlb_out$ATOXGR <- as.factor(adlb_out$ATOXGR)
  adlb_out$BTOXGR <- as.factor(adlb_out$BTOXGR)

  formatters::var_labels(adlb_out) <- adlb_var_labels

  adlb_out
}

#' Count patients with marked laboratory abnormalities
#'
#' @description `r lifecycle::badge("stable")`
#'
#' The analyze function [count_abnormal_by_marked()] creates a layout element to count patients with marked laboratory
#' abnormalities for each direction of abnormality, categorized by parameter value.
#'
#' This function analyzes primary analysis variable `var` which indicates whether a single, replicated,
#' or last marked laboratory abnormality was observed. Levels of `var` to include for each marked lab
#' abnormality (`single` and `last_replicated`) can be supplied via the `category` parameter. Additional
#' analysis variables that can be supplied as a list via the `variables` parameter are `id` (defaults
#' to `USUBJID`), a variable to indicate unique subject identifiers, `param` (defaults to `PARAM`), a
#' variable to indicate parameter values, and `direction` (defaults to `abn_dir`), a variable to indicate
#' abnormality directions.
#'
#' For each combination of `param` and `direction` levels, marked lab abnormality counts are calculated
#' as follows:
#'   * `Single, not last` & `Last or replicated`: The number of patients with `Single, not last`
#'     and `Last or replicated` values, respectively.
#'   * `Any`: The number of patients with either single or replicated marked abnormalities.
#'
#' Fractions are calculated by dividing the above counts by the number of patients with at least one
#' valid measurement recorded during the analysis.
#'
#' Prior to using this function in your table layout you must use [rtables::split_rows_by()] to create two
#' row splits, one on variable `param` and one on variable `direction`.
#'
#' @inheritParams argument_convention
#' @param category (`list`)\cr a list with different marked category names for single
#'   and last or replicated.
#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("abnormal_by_marked")`
#'   to see available statistics for this function.
#'
#' @note `Single, not last` and `Last or replicated` levels are mutually exclusive. If a patient has
#'   abnormalities that meet both the `Single, not last` and `Last or replicated` criteria, then the
#'   patient will be counted only under the `Last or replicated` category.
#'
#' @name abnormal_by_marked
#' @order 1
NULL

#' @describeIn abnormal_by_marked Statistics function for patients with marked lab abnormalities.
#'
#' @return
#' * `s_count_abnormal_by_marked()` returns statistic `count_fraction` with `Single, not last`,
#'   `Last or replicated`, and `Any` results.
#'
#' @keywords internal
s_count_abnormal_by_marked <- function(df,
                                       .var = "AVALCAT1",
                                       .spl_context,
                                       category = list(single = "SINGLE", last_replicated = c("LAST", "REPLICATED")),
                                       variables = list(id = "USUBJID", param = "PARAM", direction = "abn_dir")) {
  checkmate::assert_string(.var)
  checkmate::assert_list(variables)
  checkmate::assert_list(category)
  checkmate::assert_subset(names(category), c("single", "last_replicated"))
  checkmate::assert_subset(names(variables), c("id", "param", "direction"))
  checkmate::assert_vector(unique(df[[variables$direction]]), max.len = 1)

  assert_df_with_variables(df, c(aval = .var, variables))
  checkmate::assert_multi_class(df[[.var]], classes = c("factor", "character"))
  checkmate::assert_multi_class(df[[variables$id]], classes = c("factor", "character"))


  first_row <- .spl_context[.spl_context$split == variables[["param"]], ]
  # Patients in the denominator have at least one post-baseline visit.
  subj <- first_row$full_parent_df[[1]][[variables[["id"]]]]
  subj_cur_col <- subj[first_row$cur_col_subset[[1]]]
  # Some subjects may have a record for high and low directions but
  # should be counted only once.
  denom <- length(unique(subj_cur_col))

  if (denom != 0) {
    subjects_last_replicated <- unique(
      df[df[[.var]] %in% category[["last_replicated"]], variables$id, drop = TRUE]
    )
    subjects_single <- unique(
      df[df[[.var]] %in% category[["single"]], variables$id, drop = TRUE]
    )
    # Subjects who have both single and last/replicated abnormalities are counted in only the last/replicated group.
    subjects_single <- setdiff(subjects_single, subjects_last_replicated)
    n_single <- length(subjects_single)
    n_last_replicated <- length(subjects_last_replicated)
    n_any <- n_single + n_last_replicated
    result <- list(count_fraction = list(
      "Single, not last" = c(n_single, n_single / denom),
      "Last or replicated" = c(n_last_replicated, n_last_replicated / denom),
      "Any Abnormality" = c(n_any, n_any / denom)
    ))
  } else {
    result <- list(count_fraction = list(
      "Single, not last" = c(0, 0),
      "Last or replicated" = c(0, 0),
      "Any Abnormality" = c(0, 0)
    ))
  }

  result
}

#' @describeIn abnormal_by_marked Formatted analysis function which is used as `afun`
#'   in `count_abnormal_by_marked()`.
#'
#' @return
#' * `a_count_abnormal_by_marked()` returns the corresponding list with formatted [rtables::CellValue()].
#'
#' @keywords internal
a_count_abnormal_by_marked <- make_afun(
  s_count_abnormal_by_marked,
  .formats = c(count_fraction = format_count_fraction)
)

#' @describeIn abnormal_by_marked Layout-creating function which can take statistics function arguments
#'   and additional format arguments. This function is a wrapper for [rtables::analyze()].
#'
#' @return
#' * `count_abnormal_by_marked()` returns a layout object suitable for passing to further layouting functions,
#'   or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
#'   the statistics from `s_count_abnormal_by_marked()` to the table layout.
#'
#' @examples
#' library(dplyr)
#'
#' df <- data.frame(
#'   USUBJID = as.character(c(rep(1, 5), rep(2, 5), rep(1, 5), rep(2, 5))),
#'   ARMCD = factor(c(rep("ARM A", 5), rep("ARM B", 5), rep("ARM A", 5), rep("ARM B", 5))),
#'   ANRIND = factor(c(
#'     "NORMAL", "HIGH", "HIGH", "HIGH HIGH", "HIGH",
#'     "HIGH", "HIGH", "HIGH HIGH", "NORMAL", "HIGH HIGH", "NORMAL", "LOW", "LOW", "LOW LOW", "LOW",
#'     "LOW", "LOW", "LOW LOW", "NORMAL", "LOW LOW"
#'   )),
#'   ONTRTFL = rep(c("", "Y", "Y", "Y", "Y", "Y", "Y", "Y", "Y", "Y"), 2),
#'   PARAMCD = factor(c(rep("CRP", 10), rep("ALT", 10))),
#'   AVALCAT1 = factor(rep(c("", "", "", "SINGLE", "REPLICATED", "", "", "LAST", "", "SINGLE"), 2)),
#'   stringsAsFactors = FALSE
#' )
#'
#' df <- df %>%
#'   mutate(abn_dir = factor(
#'     case_when(
#'       ANRIND == "LOW LOW" ~ "Low",
#'       ANRIND == "HIGH HIGH" ~ "High",
#'       TRUE ~ ""
#'     ),
#'     levels = c("Low", "High")
#'   ))
#'
#' # Select only post-baseline records.
#' df <- df %>% filter(ONTRTFL == "Y")
#' df_crp <- df %>%
#'   filter(PARAMCD == "CRP") %>%
#'   droplevels()
#' full_parent_df <- list(df_crp, "not_needed")
#' cur_col_subset <- list(rep(TRUE, nrow(df_crp)), "not_needed")
#' spl_context <- data.frame(
#'   split = c("PARAMCD", "GRADE_DIR"),
#'   full_parent_df = I(full_parent_df),
#'   cur_col_subset = I(cur_col_subset)
#' )
#'
#' map <- unique(
#'   df[df$abn_dir %in% c("Low", "High") & df$AVALCAT1 != "", c("PARAMCD", "abn_dir")]
#' ) %>%
#'   lapply(as.character) %>%
#'   as.data.frame() %>%
#'   arrange(PARAMCD, abn_dir)
#'
#' basic_table() %>%
#'   split_cols_by("ARMCD") %>%
#'   split_rows_by("PARAMCD") %>%
#'   summarize_num_patients(
#'     var = "USUBJID",
#'     .stats = "unique_count"
#'   ) %>%
#'   split_rows_by(
#'     "abn_dir",
#'     split_fun = trim_levels_to_map(map)
#'   ) %>%
#'   count_abnormal_by_marked(
#'     var = "AVALCAT1",
#'     variables = list(
#'       id = "USUBJID",
#'       param = "PARAMCD",
#'       direction = "abn_dir"
#'     )
#'   ) %>%
#'   build_table(df = df)
#'
#' basic_table() %>%
#'   split_cols_by("ARMCD") %>%
#'   split_rows_by("PARAMCD") %>%
#'   summarize_num_patients(
#'     var = "USUBJID",
#'     .stats = "unique_count"
#'   ) %>%
#'   split_rows_by(
#'     "abn_dir",
#'     split_fun = trim_levels_in_group("abn_dir")
#'   ) %>%
#'   count_abnormal_by_marked(
#'     var = "AVALCAT1",
#'     variables = list(
#'       id = "USUBJID",
#'       param = "PARAMCD",
#'       direction = "abn_dir"
#'     )
#'   ) %>%
#'   build_table(df = df)
#'
#' @export
#' @order 2
count_abnormal_by_marked <- function(lyt,
                                     var,
                                     category = list(single = "SINGLE", last_replicated = c("LAST", "REPLICATED")),
                                     variables = list(id = "USUBJID", param = "PARAM", direction = "abn_dir"),
                                     na_str = default_na_str(),
                                     nested = TRUE,
                                     ...,
                                     .stats = NULL,
                                     .formats = NULL,
                                     .labels = NULL,
                                     .indent_mods = NULL) {
  checkmate::assert_string(var)

  extra_args <- list(category = category, variables = variables, ...)

  afun <- make_afun(
    a_count_abnormal_by_marked,
    .stats = .stats,
    .formats = .formats,
    .labels = .labels,
    .indent_mods = .indent_mods,
    .ungroup_stats = "count_fraction"
  )

  lyt <- analyze(
    lyt = lyt,
    vars = var,
    afun = afun,
    na_str = na_str,
    nested = nested,
    show_labels = "hidden",
    extra_args = extra_args
  )
  lyt
}

#' Formatting functions
#'
#' @description `r lifecycle::badge("stable")`
#'
#' See below for the list of formatting functions created in `tern` to work with `rtables`.
#'
#' Other available formats can be listed via [`formatters::list_valid_format_labels()`]. Additional
#' custom formats can be created via the [`formatters::sprintf_format()`] function.
#'
#' @family formatting functions
#' @name formatting_functions
NULL

#' Format fraction and percentage
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Formats a fraction together with ratio in percent.
#'
#' @param x (named `integer`)\cr vector with elements `num` and `denom`.
#' @param ... not used. Required for `rtables` interface.
#'
#' @return A string in the format `num / denom (ratio %)`. If `num` is 0, the format is `num / denom`.
#'
#' @examples
#' format_fraction(x = c(num = 2L, denom = 3L))
#' format_fraction(x = c(num = 0L, denom = 3L))
#'
#' @family formatting functions
#' @export
format_fraction <- function(x, ...) {
  attr(x, "label") <- NULL

  checkmate::assert_vector(x)
  checkmate::assert_count(x["num"])
  checkmate::assert_count(x["denom"])

  result <- if (x["num"] == 0) {
    paste0(x["num"], "/", x["denom"])
  } else {
    paste0(
      x["num"], "/", x["denom"],
      " (", round(x["num"] / x["denom"] * 100, 1), "%)"
    )
  }

  return(result)
}

#' Format fraction and percentage with fixed single decimal place
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Formats a fraction together with ratio in percent with fixed single decimal place.
#' Includes trailing zero in case of whole number percentages to always keep one decimal place.
#'
#' @inheritParams format_fraction
#'
#' @return A string in the format `num / denom (ratio %)`. If `num` is 0, the format is `num / denom`.
#'
#' @examples
#' format_fraction_fixed_dp(x = c(num = 1L, denom = 2L))
#' format_fraction_fixed_dp(x = c(num = 1L, denom = 4L))
#' format_fraction_fixed_dp(x = c(num = 0L, denom = 3L))
#'
#' @family formatting functions
#' @export
format_fraction_fixed_dp <- function(x, ...) {
  attr(x, "label") <- NULL
  checkmate::assert_vector(x)
  checkmate::assert_count(x["num"])
  checkmate::assert_count(x["denom"])

  result <- if (x["num"] == 0) {
    paste0(x["num"], "/", x["denom"])
  } else {
    paste0(
      x["num"], "/", x["denom"],
      " (", sprintf("%.1f", round(x["num"] / x["denom"] * 100, 1)), "%)"
    )
  }
  return(result)
}

#' Format count and fraction
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Formats a count together with fraction with special consideration when count is `0`.
#'
#' @param x (`numeric(2)`)\cr vector of length 2 with count and fraction, respectively.
#' @param ... not used. Required for `rtables` interface.
#'
#' @return A string in the format `count (fraction %)`. If `count` is 0, the format is `0`.
#'
#' @examples
#' format_count_fraction(x = c(2, 0.6667))
#' format_count_fraction(x = c(0, 0))
#'
#' @family formatting functions
#' @export
format_count_fraction <- function(x, ...) {
  attr(x, "label") <- NULL

  if (any(is.na(x))) {
    return("NA")
  }

  checkmate::assert_vector(x)
  checkmate::assert_integerish(x[1])
  assert_proportion_value(x[2], include_boundaries = TRUE)

  result <- if (x[1] == 0) {
    "0"
  } else {
    paste0(x[1], " (", round(x[2] * 100, 1), "%)")
  }

  return(result)
}

#' Format count and percentage with fixed single decimal place
#'
#' @description `r lifecycle::badge("experimental")`
#'
#' Formats a count together with fraction with special consideration when count is `0`.
#'
#' @inheritParams format_count_fraction
#'
#' @return A string in the format `count (fraction %)`. If `count` is 0, the format is `0`.
#'
#' @examples
#' format_count_fraction_fixed_dp(x = c(2, 0.6667))
#' format_count_fraction_fixed_dp(x = c(2, 0.5))
#' format_count_fraction_fixed_dp(x = c(0, 0))
#'
#' @family formatting functions
#' @export
format_count_fraction_fixed_dp <- function(x, ...) {
  attr(x, "label") <- NULL

  if (any(is.na(x))) {
    return("NA")
  }

  checkmate::assert_vector(x)
  checkmate::assert_integerish(x[1])
  assert_proportion_value(x[2], include_boundaries = TRUE)

  result <- if (x[1] == 0) {
    "0"
  } else if (.is_equal_float(x[2], 1)) {
    sprintf("%d (100%%)", x[1])
  } else {
    sprintf("%d (%.1f%%)", x[1], x[2] * 100)
  }

  return(result)
}

#' Format count and fraction with special case for count < 10
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Formats a count together with fraction with special consideration when count is less than 10.
#'
#' @inheritParams format_count_fraction
#'
#' @return A string in the format `count (fraction %)`. If `count` is less than 10, only `count` is printed.
#'
#' @examples
#' format_count_fraction_lt10(x = c(275, 0.9673))
#' format_count_fraction_lt10(x = c(2, 0.6667))
#' format_count_fraction_lt10(x = c(9, 1))
#'
#' @family formatting functions
#' @export
format_count_fraction_lt10 <- function(x, ...) {
  attr(x, "label") <- NULL

  if (any(is.na(x))) {
    return("NA")
  }

  checkmate::assert_vector(x)
  checkmate::assert_integerish(x[1])
  assert_proportion_value(x[2], include_boundaries = TRUE)

  result <- if (x[1] < 10) {
    paste0(x[1])
  } else {
    paste0(x[1], " (", round(x[2] * 100, 1), "%)")
  }

  return(result)
}

#' Format XX as a formatting function
#'
#' Translate a string where x and dots are interpreted as number place
#' holders, and others as formatting elements.
#'
#' @param str (`string`)\cr template.
#'
#' @return An `rtables` formatting function.
#'
#' @examples
#' test <- list(c(1.658, 0.5761), c(1e1, 785.6))
#'
#' z <- format_xx("xx (xx.x)")
#' sapply(test, z)
#'
#' z <- format_xx("xx.x - xx.x")
#' sapply(test, z)
#'
#' z <- format_xx("xx.x, incl. xx.x% NE")
#' sapply(test, z)
#'
#' @family formatting functions
#' @export
format_xx <- function(str) {
  # Find position in the string.
  positions <- gregexpr(pattern = "x+\\.x+|x+", text = str, perl = TRUE)
  x_positions <- regmatches(x = str, m = positions)[[1]]

  # Roundings depends on the number of x behind [.].
  roundings <- lapply(
    X = x_positions,
    function(x) {
      y <- strsplit(split = "\\.", x = x)[[1]]
      rounding <- function(x) {
        round(x, digits = ifelse(length(y) > 1, nchar(y[2]), 0))
      }
      return(rounding)
    }
  )

  rtable_format <- function(x, output) {
    values <- Map(y = x, fun = roundings, function(y, fun) fun(y))
    regmatches(x = str, m = positions)[[1]] <- values
    return(str)
  }

  return(rtable_format)
}

#' Format numeric values by significant figures
#'
#' Format numeric values to print with a specified number of significant figures.
#'
#' @param sigfig (`integer(1)`)\cr number of significant figures to display.
#' @param format (`string`)\cr the format label (string) to apply when printing the value. Decimal
#'   places in string are ignored in favor of formatting by significant figures. Formats options are:
#'   `"xx"`, `"xx / xx"`, `"(xx, xx)"`, `"xx - xx"`, and `"xx (xx)"`.
#' @param num_fmt (`string`)\cr numeric format modifiers to apply to the value. Defaults to `"fg"` for
#'   standard significant figures formatting - fixed (non-scientific notation) format (`"f"`)
#'   and `sigfig` equal to number of significant figures instead of decimal places (`"g"`). See the
#'   [formatC()] `format` argument for more options.
#'
#' @return An `rtables` formatting function.
#'
#' @examples
#' fmt_3sf <- format_sigfig(3)
#' fmt_3sf(1.658)
#' fmt_3sf(1e1)
#'
#' fmt_5sf <- format_sigfig(5)
#' fmt_5sf(0.57)
#' fmt_5sf(0.000025645)
#'
#' @family formatting functions
#' @export
format_sigfig <- function(sigfig, format = "xx", num_fmt = "fg") {
  checkmate::assert_integerish(sigfig)
  format <- gsub("xx\\.|xx\\.x+", "xx", format)
  checkmate::assert_choice(format, c("xx", "xx / xx", "(xx, xx)", "xx - xx", "xx (xx)"))
  function(x, ...) {
    if (!is.numeric(x)) stop("`format_sigfig` cannot be used for non-numeric values. Please choose another format.")
    num <- formatC(signif(x, digits = sigfig), digits = sigfig, format = num_fmt, flag = "#")
    num <- gsub("\\.$", "", num) # remove trailing "."

    format_value(num, format)
  }
}

#' Format fraction with lower threshold
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Formats a fraction when the second element of the input `x` is the fraction. It applies
#' a lower threshold, below which it is just stated that the fraction is smaller than that.
#'
#' @param threshold (`proportion`)\cr lower threshold.
#'
#' @return An `rtables` formatting function that takes numeric input `x` where the second
#'   element is the fraction that is formatted. If the fraction is above or equal to the threshold,
#'   then it is displayed in percentage. If it is positive but below the threshold, it returns,
#'   e.g. "<1" if the threshold is `0.01`. If it is zero, then just "0" is returned.
#'
#' @examples
#' format_fun <- format_fraction_threshold(0.05)
#' format_fun(x = c(20, 0.1))
#' format_fun(x = c(2, 0.01))
#' format_fun(x = c(0, 0))
#'
#' @family formatting functions
#' @export
format_fraction_threshold <- function(threshold) {
  assert_proportion_value(threshold)
  string_below_threshold <- paste0("<", round(threshold * 100))
  function(x, ...) {
    assert_proportion_value(x[2], include_boundaries = TRUE)
    ifelse(
      x[2] > 0.01,
      round(x[2] * 100),
      ifelse(
        x[2] == 0,
        "0",
        string_below_threshold
      )
    )
  }
}

#' Format extreme values
#'
#' @description `r lifecycle::badge("stable")`
#'
#' `rtables` formatting functions that handle extreme values.
#'
#' @param digits (`integer(1)`)\cr number of decimal places to display.
#'
#' @details For each input, apply a format to the specified number of `digits`. If the value is
#'    below a threshold, it returns "<0.01" e.g. if the number of `digits` is 2. If the value is
#'    above a threshold, it returns ">999.99" e.g. if the number of `digits` is 2.
#'    If it is zero, then returns "0.00".
#'
#' @family formatting functions
#' @name extreme_format
NULL

#' @describeIn extreme_format Internal helper function to calculate the threshold and create formatted strings
#'  used in Formatting Functions. Returns a list with elements `threshold` and `format_string`.
#'
#' @return
#' * `h_get_format_threshold()` returns a `list` of 2 elements: `threshold`, with `low` and `high` thresholds,
#'   and `format_string`, with thresholds formatted as strings.
#'
#' @examples
#' h_get_format_threshold(2L)
#'
#' @export
h_get_format_threshold <- function(digits = 2L) {
  checkmate::assert_integerish(digits)

  low_threshold <- 1 / (10 ^ digits) # styler: off
  high_threshold <- 1000 - (1 / (10 ^ digits)) # styler: off

  string_below_threshold <- paste0("<", low_threshold)
  string_above_threshold <- paste0(">", high_threshold)

  list(
    "threshold" = c(low = low_threshold, high = high_threshold),
    "format_string" = c(low = string_below_threshold, high = string_above_threshold)
  )
}

#' @describeIn extreme_format Internal helper function to apply a threshold format to a value.
#'   Creates a formatted string to be used in Formatting Functions.
#'
#' @param x (`numeric(1)`)\cr value to format.
#'
#' @return
#' * `h_format_threshold()` returns the given value, or if the value is not within the digit threshold the relation
#'   of the given value to the digit threshold, as a formatted string.
#'
#' @examples
#' h_format_threshold(0.001)
#' h_format_threshold(1000)
#'
#' @export
h_format_threshold <- function(x, digits = 2L) {
  if (is.na(x)) {
    return(x)
  }

  checkmate::assert_numeric(x, lower = 0)

  l_fmt <- h_get_format_threshold(digits)

  result <- if (x < l_fmt$threshold["low"] && 0 < x) {
    l_fmt$format_string["low"]
  } else if (x > l_fmt$threshold["high"]) {
    l_fmt$format_string["high"]
  } else {
    sprintf(fmt = paste0("%.", digits, "f"), x)
  }

  unname(result)
}

#' Format a single extreme value
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Create a formatting function for a single extreme value.
#'
#' @inheritParams extreme_format
#'
#' @return An `rtables` formatting function that uses threshold `digits` to return a formatted extreme value.
#'
#' @examples
#' format_fun <- format_extreme_values(2L)
#' format_fun(x = 0.127)
#' format_fun(x = Inf)
#' format_fun(x = 0)
#' format_fun(x = 0.009)
#'
#' @family formatting functions
#' @export
format_extreme_values <- function(digits = 2L) {
  function(x, ...) {
    checkmate::assert_scalar(x, na.ok = TRUE)

    h_format_threshold(x = x, digits = digits)
  }
}

#' Format extreme values part of a confidence interval
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Formatting Function for extreme values part of a confidence interval. Values
#' are formatted as e.g. "(xx.xx, xx.xx)" if the number of `digits` is 2.
#'
#' @inheritParams extreme_format
#'
#' @return An `rtables` formatting function that uses threshold `digits` to return a formatted extreme
#'   values confidence interval.
#'
#' @examples
#' format_fun <- format_extreme_values_ci(2L)
#' format_fun(x = c(0.127, Inf))
#' format_fun(x = c(0, 0.009))
#'
#' @family formatting functions
#' @export
format_extreme_values_ci <- function(digits = 2L) {
  function(x, ...) {
    checkmate::assert_vector(x, len = 2)
    l_result <- h_format_threshold(x = x[1], digits = digits)
    h_result <- h_format_threshold(x = x[2], digits = digits)

    paste0("(", l_result, ", ", h_result, ")")
  }
}

#' Format automatically using data significant digits
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Formatting function for the majority of default methods used in [analyze_vars()].
#' For non-derived values, the significant digits of data is used (e.g. range), while derived
#' values have one more digits (measure of location and dispersion like mean, standard deviation).
#' This function can be called internally with "auto" like, for example,
#' `.formats = c("mean" = "auto")`. See details to see how this works with the inner function.
#'
#' @param dt_var (`numeric`)\cr variable data the statistics were calculated from. Used only to
#'   find significant digits. In [analyze_vars] this comes from `.df_row` (see
#'   [rtables::additional_fun_params]), and it is the row data after the above row splits. No
#'   column split is considered.
#' @param x_stat (`string`)\cr string indicating the current statistical method used.
#'
#' @return A string that `rtables` prints in a table cell.
#'
#' @details
#' The internal function is needed to work with `rtables` default structure for
#' format functions, i.e. `function(x, ...)`, where is x are results from statistical evaluation.
#' It can be more than one element (e.g. for `.stats = "mean_sd"`).
#'
#' @examples
#' x_todo <- c(0.001, 0.2, 0.0011000, 3, 4)
#' res <- c(mean(x_todo[1:3]), sd(x_todo[1:3]))
#'
#' # x is the result coming into the formatting function -> res!!
#' format_auto(dt_var = x_todo, x_stat = "mean_sd")(x = res)
#' format_auto(x_todo, "range")(x = range(x_todo))
#' no_sc_x <- c(0.0000001, 1)
#' format_auto(no_sc_x, "range")(x = no_sc_x)
#'
#' @family formatting functions
#' @export
format_auto <- function(dt_var, x_stat) {
  function(x = "", ...) {
    checkmate::assert_numeric(x, min.len = 1)
    checkmate::assert_numeric(dt_var, min.len = 1)
    # Defaults - they may be a param in the future
    der_stats <- c(
      "mean", "sd", "se", "median", "geom_mean", "quantiles", "iqr",
      "mean_sd", "mean_se", "mean_se", "mean_ci", "mean_sei", "mean_sdi",
      "median_ci"
    )
    nonder_stats <- c("n", "range", "min", "max")

    # Safenet for miss-modifications
    stopifnot(length(intersect(der_stats, nonder_stats)) == 0) # nolint
    checkmate::assert_choice(x_stat, c(der_stats, nonder_stats))

    # Finds the max number of digits in data
    detect_dig <- vapply(dt_var, count_decimalplaces, FUN.VALUE = numeric(1)) %>%
      max()

    if (x_stat %in% der_stats) {
      detect_dig <- detect_dig + 1
    }

    # Render input
    str_vals <- formatC(x, digits = detect_dig, format = "f")
    def_fmt <- get_formats_from_stats(x_stat)[[x_stat]]
    str_fmt <- str_extract(def_fmt, invert = FALSE)[[1]]
    if (length(str_fmt) != length(str_vals)) {
      stop(
        "Number of inserted values as result (", length(str_vals),
        ") is not the same as there should be in the default tern formats for ",
        x_stat, " (-> ", def_fmt, " needs ", length(str_fmt), " values). ",
        "See tern_default_formats to check all of them."
      )
    }

    # Squashing them together
    inv_str_fmt <- str_extract(def_fmt, invert = TRUE)[[1]]
    stopifnot(length(inv_str_fmt) == length(str_vals) + 1) # nolint

    out <- vector("character", length = length(inv_str_fmt) + length(str_vals))
    is_even <- seq_along(out) %% 2 == 0
    out[is_even] <- str_vals
    out[!is_even] <- inv_str_fmt

    return(paste0(out, collapse = ""))
  }
}

# Utility function that could be useful in general
str_extract <- function(string, pattern = "xx|xx\\.|xx\\.x+", invert = FALSE) {
  regmatches(string, gregexpr(pattern, string), invert = invert)
}

# Helper function
count_decimalplaces <- function(dec) {
  if (is.na(dec)) {
    return(0)
  } else if (abs(dec - round(dec)) > .Machine$double.eps^0.5) { # For precision
    nchar(strsplit(format(dec, scientific = FALSE, trim = FALSE), ".", fixed = TRUE)[[1]][[2]])
  } else {
    return(0)
  }
}

#' Apply automatic formatting
#'
#' Checks if any of the listed formats in `.formats` are `"auto"`, and replaces `"auto"` with
#' the correct implementation of `format_auto` for the given statistics, data, and variable.
#'
#' @inheritParams argument_convention
#' @param x_stats (named `list`)\cr a named list of statistics where each element corresponds
#'   to an element in `.formats`, with matching names.
#'
#' @keywords internal
apply_auto_formatting <- function(.formats, x_stats, .df_row, .var) {
  is_auto_fmt <- vapply(.formats, function(ii) is.character(ii) && ii == "auto", logical(1))
  if (any(is_auto_fmt)) {
    auto_stats <- x_stats[is_auto_fmt]
    var_df <- .df_row[[.var]] # xxx this can be extended for the WHOLE data or single facets
    .formats[is_auto_fmt] <- lapply(names(auto_stats), format_auto, dt_var = var_df)
  }
  .formats
}

#' Univariate formula special term
#'
#' @description `r lifecycle::badge("stable")`
#'
#' The special term `univariate` indicate that the model should be fitted individually for
#' every variable included in univariate.
#'
#' @param x (`character`)\cr a vector of variable names separated by commas.
#'
#' @return When used within a model formula, produces univariate models for each variable provided.
#'
#' @details
#' If provided alongside with pairwise specification, the model
#' `y ~ ARM + univariate(SEX, AGE, RACE)` lead to the study and comparison of the models
#' + `y ~ ARM`
#' + `y ~ ARM + SEX`
#' + `y ~ ARM + AGE`
#' + `y ~ ARM + RACE`
#'
#' @export
univariate <- function(x) {
  structure(x, varname = deparse(substitute(x)))
}

# Get the right-hand-term of a formula
rht <- function(x) {
  checkmate::assert_formula(x)
  y <- as.character(rev(x)[[1]])
  return(y)
}

#' Hazard ratio estimation in interactions
#'
#' This function estimates the hazard ratios between arms when an interaction variable is given with
#' specific values.
#'
#' @param variable,given (`character(2)`)\cr names of the two variables in the interaction. We seek the estimation of
#'   the levels of `variable` given the levels of `given`.
#' @param lvl_var,lvl_given (`character`)\cr corresponding levels given by [levels()].
#' @param mmat (named `numeric`) a vector filled with `0`s used as a template to obtain the design matrix.
#' @param coef (`numeric`)\cr vector of estimated coefficients.
#' @param vcov (`matrix`)\cr variance-covariance matrix of underlying model.
#' @param conf_level (`proportion`)\cr confidence level of estimate intervals.
#'
#' @details Given the cox regression investigating the effect of Arm (A, B, C; reference A)
#'   and Sex (F, M; reference Female). The model is abbreviated: y ~ Arm + Sex + Arm x Sex.
#'   The cox regression estimates the coefficients along with a variance-covariance matrix for:
#'
#'   - b1 (arm b), b2 (arm c)
#'   - b3 (sex m)
#'   - b4 (arm b: sex m), b5 (arm c: sex m)
#'
#'   Given that I want an estimation of the Hazard Ratio for arm C/sex M, the estimation
#'   will be given in reference to arm A/Sex M by exp(b2 + b3 + b5)/ exp(b3) = exp(b2 + b5),
#'   therefore the interaction coefficient is given by b2 + b5 while the standard error is obtained
#'   as $1.96 * sqrt(Var b2 + Var b5 + 2 * covariance (b2,b5))$ for a confidence level of 0.95.
#'
#' @return A list of matrices (one per level of variable) with rows corresponding to the combinations of
#'   `variable` and `given`, with columns:
#'   * `coef_hat`: Estimation of the coefficient.
#'   * `coef_se`: Standard error of the estimation.
#'   * `hr`: Hazard ratio.
#'   * `lcl, ucl`: Lower/upper confidence limit of the hazard ratio.
#'
#' @seealso [s_cox_multivariate()].
#'
#' @examples
#' library(dplyr)
#' library(survival)
#'
#' ADSL <- tern_ex_adsl %>%
#'   filter(SEX %in% c("F", "M"))
#'
#' adtte <- tern_ex_adtte %>% filter(PARAMCD == "PFS")
#' adtte$ARMCD <- droplevels(adtte$ARMCD)
#' adtte$SEX <- droplevels(adtte$SEX)
#'
#' mod <- coxph(
#'   formula = Surv(time = AVAL, event = 1 - CNSR) ~ (SEX + ARMCD)^2,
#'   data = adtte
#' )
#'
#' mmat <- stats::model.matrix(mod)[1, ]
#' mmat[!mmat == 0] <- 0
#'
#' @keywords internal
estimate_coef <- function(variable, given,
                          lvl_var, lvl_given,
                          coef,
                          mmat,
                          vcov,
                          conf_level = 0.95) {
  var_lvl <- paste0(variable, lvl_var[-1]) # [-1]: reference level
  giv_lvl <- paste0(given, lvl_given)

  design_mat <- expand.grid(variable = var_lvl, given = giv_lvl)
  design_mat <- design_mat[order(design_mat$variable, design_mat$given), ]
  design_mat <- within(
    data = design_mat,
    expr = {
      inter <- paste0(variable, ":", given)
      rev_inter <- paste0(given, ":", variable)
    }
  )

  split_by_variable <- design_mat$variable
  interaction_names <- paste(design_mat$variable, design_mat$given, sep = "/")

  design_mat <- apply(
    X = design_mat, MARGIN = 1, FUN = function(x) {
      mmat[names(mmat) %in% x[-which(names(x) == "given")]] <- 1
      return(mmat)
    }
  )
  colnames(design_mat) <- interaction_names

  betas <- as.matrix(coef)

  coef_hat <- t(design_mat) %*% betas
  dimnames(coef_hat)[2] <- "coef"

  coef_se <- apply(design_mat, 2, function(x) {
    vcov_el <- as.logical(x)
    y <- vcov[vcov_el, vcov_el]
    y <- sum(y)
    y <- sqrt(y)
    return(y)
  })

  q_norm <- stats::qnorm((1 + conf_level) / 2)
  y <- cbind(coef_hat, `se(coef)` = coef_se)

  y <- apply(y, 1, function(x) {
    x["hr"] <- exp(x["coef"])
    x["lcl"] <- exp(x["coef"] - q_norm * x["se(coef)"])
    x["ucl"] <- exp(x["coef"] + q_norm * x["se(coef)"])

    return(x)
  })

  y <- t(y)
  y <- by(y, split_by_variable, identity)
  y <- lapply(y, as.matrix)

  attr(y, "details") <- paste0(
    "Estimations of ", variable,
    " hazard ratio given the level of ", given, " compared to ",
    variable, " level ", lvl_var[1], "."
  )
  return(y)
}

#' `tryCatch` around `car::Anova`
#'
#' Captures warnings when executing [car::Anova].
#'
#' @inheritParams car::Anova
#'
#' @return A list with item `aov` for the result of the model and `error_text` for the captured warnings.
#'
#' @examples
#' # `car::Anova` on cox regression model including strata and expected
#' # a likelihood ratio test triggers a warning as only Wald method is
#' # accepted.
#'
#' library(survival)
#'
#' mod <- coxph(
#'   formula = Surv(time = futime, event = fustat) ~ factor(rx) + strata(ecog.ps),
#'   data = ovarian
#' )
#'
#' @keywords internal
try_car_anova <- function(mod,
                          test.statistic) { # nolint
  y <- tryCatch(
    withCallingHandlers(
      expr = {
        warn_text <- c()
        list(
          aov = car::Anova(
            mod,
            test.statistic = test.statistic,
            type = "III"
          ),
          warn_text = warn_text
        )
      },
      warning = function(w) {
        # If a warning is detected it is handled as "w".
        warn_text <<- trimws(paste0("Warning in `try_car_anova`: ", w))

        # A warning is sometimes expected, then, we want to restart
        # the execution while ignoring the warning.
        invokeRestart("muffleWarning")
      }
    ),
    finally = {
    }
  )

  return(y)
}

#' Fit a Cox regression model and ANOVA
#'
#' The functions derives the effect p-values using [car::Anova()] from [survival::coxph()] results.
#'
#' @inheritParams t_coxreg
#'
#' @return A list with items `mod` (results of [survival::coxph()]), `msum` (result of `summary`) and
#'   `aov` (result of [car::Anova()]).
#'
#' @noRd
fit_n_aov <- function(formula,
                      data = data,
                      conf_level = conf_level,
                      pval_method = c("wald", "likelihood"),
                      ...) {
  pval_method <- match.arg(pval_method)

  environment(formula) <- environment()
  suppressWarnings({
    # We expect some warnings due to coxph which fails strict programming.
    mod <- survival::coxph(formula, data = data, ...)
    msum <- summary(mod, conf.int = conf_level)
  })

  aov <- try_car_anova(
    mod,
    test.statistic = switch(pval_method,
      "wald" = "Wald",
      "likelihood" = "LR"
    )
  )

  warn_attr <- aov$warn_text
  if (!is.null(aov$warn_text)) message(warn_attr)

  aov <- aov$aov
  y <- list(mod = mod, msum = msum, aov = aov)
  attr(y, "message") <- warn_attr

  return(y)
}

# argument_checks
check_formula <- function(formula) {
  if (!(inherits(formula, "formula"))) {
    stop("Check `formula`. A formula should resemble `Surv(time = AVAL, event = 1 - CNSR) ~ study_arm(ARMCD)`.")
  }

  invisible()
}

check_covariate_formulas <- function(covariates) {
  if (!all(vapply(X = covariates, FUN = inherits, what = "formula", FUN.VALUE = TRUE)) || is.null(covariates)) {
    stop("Check `covariates`, it should be a list of right-hand-term formulas, e.g. list(Age = ~AGE).")
  }

  invisible()
}

name_covariate_names <- function(covariates) {
  miss_names <- names(covariates) == ""
  no_names <- is.null(names(covariates))
  if (any(miss_names)) names(covariates)[miss_names] <- vapply(covariates[miss_names], FUN = rht, FUN.VALUE = "name")
  if (no_names) names(covariates) <- vapply(covariates, FUN = rht, FUN.VALUE = "name")
  return(covariates)
}

check_increments <- function(increments, covariates) {
  if (!is.null(increments)) {
    covariates <- vapply(covariates, FUN = rht, FUN.VALUE = "name")
    lapply(
      X = names(increments), FUN = function(x) {
        if (!x %in% covariates) {
          warning(
            paste(
              "Check `increments`, the `increment` for ", x,
              "doesn't match any names in investigated covariate(s)."
            )
          )
        }
      }
    )
  }

  invisible()
}

#' Multivariate Cox model - summarized results
#'
#' Analyses based on multivariate Cox model are usually not performed for the Controlled Substance Reporting or
#' regulatory documents but serve exploratory purposes only (e.g., for publication). In practice, the model usually
#' includes only the main effects (without interaction terms). It produces the hazard ratio estimates for each of the
#' covariates included in the model.
#' The analysis follows the same principles (e.g., stratified vs. unstratified analysis and tie handling) as the
#' usual Cox model analysis. Since there is usually no pre-specified hypothesis testing for such analysis,
#' the p.values need to be interpreted with caution. (**Statistical Analysis of Clinical Trials Data with R**,
#' `NEST's bookdown`)
#'
#' @param formula (`formula`)\cr a formula corresponding to the investigated [survival::Surv()] survival model
#'   including covariates.
#' @param data (`data.frame`)\cr a data frame which includes the variable in formula and covariates.
#' @param conf_level (`proportion`)\cr the confidence level for the hazard ratio interval estimations. Default is 0.95.
#' @param pval_method (`string`)\cr the method used for the estimation of p-values, should be one of
#'   `"wald"` (default) or `"likelihood"`.
#' @param ... optional parameters passed to [survival::coxph()]. Can include `ties`, a character string specifying the
#'   method for tie handling, one of `exact` (default), `efron`, `breslow`.
#'
#' @return A `list` with elements `mod`, `msum`, `aov`, and `coef_inter`.
#'
#' @details The output is limited to single effect terms. Work in ongoing for estimation of interaction terms
#'   but is out of scope as defined by the  Global Data Standards Repository
#'   (**`GDS_Standard_TLG_Specs_Tables_2.doc`**).
#'
#' @seealso [estimate_coef()].
#'
#' @examples
#' library(dplyr)
#'
#' adtte <- tern_ex_adtte
#' adtte_f <- subset(adtte, PARAMCD == "OS") # _f: filtered
#' adtte_f <- filter(
#'   adtte_f,
#'   PARAMCD == "OS" &
#'     SEX %in% c("F", "M") &
#'     RACE %in% c("ASIAN", "BLACK OR AFRICAN AMERICAN", "WHITE")
#' )
#' adtte_f$SEX <- droplevels(adtte_f$SEX)
#' adtte_f$RACE <- droplevels(adtte_f$RACE)
#'
#' @keywords internal
s_cox_multivariate <- function(formula, data,
                               conf_level = 0.95,
                               pval_method = c("wald", "likelihood"),
                               ...) {
  tf <- stats::terms(formula, specials = c("strata"))
  covariates <- rownames(attr(tf, "factors"))[-c(1, unlist(attr(tf, "specials")))]
  lapply(
    X = covariates,
    FUN = function(x) {
      if (is.character(data[[x]])) {
        data[[x]] <<- as.factor(data[[x]])
      }
      invisible()
    }
  )
  pval_method <- match.arg(pval_method)

  # Results directly exported from environment(fit_n_aov) to environment(s_function_draft)
  y <- fit_n_aov(
    formula = formula,
    data = data,
    conf_level = conf_level,
    pval_method = pval_method,
    ...
  )
  mod <- y$mod
  aov <- y$aov
  msum <- y$msum
  list2env(as.list(y), environment())

  all_term_labs <- attr(mod$terms, "term.labels")
  term_labs <- all_term_labs[which(attr(mod$terms, "order") == 1)]
  names(term_labs) <- term_labs

  coef_inter <- NULL
  if (any(attr(mod$terms, "order") > 1)) {
    for_inter <- all_term_labs[attr(mod$terms, "order") > 1]
    names(for_inter) <- for_inter
    mmat <- stats::model.matrix(mod)[1, ]
    mmat[!mmat == 0] <- 0
    mcoef <- stats::coef(mod)
    mvcov <- stats::vcov(mod)

    estimate_coef_local <- function(variable, given) {
      estimate_coef(
        variable, given,
        coef = mcoef, mmat = mmat, vcov = mvcov, conf_level = conf_level,
        lvl_var = levels(data[[variable]]), lvl_given = levels(data[[given]])
      )
    }

    coef_inter <- lapply(
      for_inter, function(x) {
        y <- attr(mod$terms, "factors")[, x]
        y <- names(y[y > 0])
        Map(estimate_coef_local, variable = y, given = rev(y))
      }
    )
  }

  list(mod = mod, msum = msum, aov = aov, coef_inter = coef_inter)
}

#' Proportion estimation
#'
#' @description `r lifecycle::badge("stable")`
#'
#' The analyze function [estimate_proportion()] creates a layout element to estimate the proportion of responders
#' within a studied population. The primary analysis variable, `vars`, indicates whether a response has occurred for
#' each record. See the `method` parameter for options of methods to use when constructing the confidence interval of
#' the proportion. Additionally, a stratification variable can be supplied via the `strata` element of the `variables`
#' argument.
#'
#' @inheritParams prop_strat_wilson
#' @inheritParams argument_convention
#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("estimate_proportion")`
#'   to see available statistics for this function.
#' @param method (`string`)\cr the method used to construct the confidence interval
#'   for proportion of successful outcomes; one of `waldcc`, `wald`, `clopper-pearson`,
#'   `wilson`, `wilsonc`, `strat_wilson`, `strat_wilsonc`, `agresti-coull` or `jeffreys`.
#' @param long (`flag`)\cr whether a long description is required.
#'
#' @seealso [h_proportions]
#'
#' @name estimate_proportion
#' @order 1
NULL

#' @describeIn estimate_proportion Statistics function estimating a
#'   proportion along with its confidence interval.
#'
#' @param df (`logical` or `data.frame`)\cr if only a logical vector is used,
#'   it indicates whether each subject is a responder or not. `TRUE` represents
#'   a successful outcome. If a `data.frame` is provided, also the `strata` variable
#'   names must be provided in `variables` as a list element with the strata strings.
#'   In the case of `data.frame`, the logical vector of responses must be indicated as a
#'   variable name in `.var`.
#'
#' @return
#' * `s_proportion()` returns statistics `n_prop` (`n` and proportion) and `prop_ci` (proportion CI) for a
#'   given variable.
#'
#' @examples
#' # Case with only logical vector.
#' rsp_v <- c(1, 0, 1, 0, 1, 1, 0, 0)
#' s_proportion(rsp_v)
#'
#' # Example for Stratified Wilson CI
#' nex <- 100 # Number of example rows
#' dta <- data.frame(
#'   "rsp" = sample(c(TRUE, FALSE), nex, TRUE),
#'   "grp" = sample(c("A", "B"), nex, TRUE),
#'   "f1" = sample(c("a1", "a2"), nex, TRUE),
#'   "f2" = sample(c("x", "y", "z"), nex, TRUE),
#'   stringsAsFactors = TRUE
#' )
#'
#' s_proportion(
#'   df = dta,
#'   .var = "rsp",
#'   variables = list(strata = c("f1", "f2")),
#'   conf_level = 0.90,
#'   method = "strat_wilson"
#' )
#'
#' @export
s_proportion <- function(df,
                         .var,
                         conf_level = 0.95,
                         method = c(
                           "waldcc", "wald", "clopper-pearson",
                           "wilson", "wilsonc", "strat_wilson", "strat_wilsonc",
                           "agresti-coull", "jeffreys"
                         ),
                         weights = NULL,
                         max_iterations = 50,
                         variables = list(strata = NULL),
                         long = FALSE) {
  method <- match.arg(method)
  checkmate::assert_flag(long)
  assert_proportion_value(conf_level)

  if (!is.null(variables$strata)) {
    # Checks for strata
    if (missing(df)) stop("When doing stratified analysis a data.frame with specific columns is needed.")
    strata_colnames <- variables$strata
    checkmate::assert_character(strata_colnames, null.ok = FALSE)
    strata_vars <- stats::setNames(as.list(strata_colnames), strata_colnames)
    assert_df_with_variables(df, strata_vars)

    strata <- interaction(df[strata_colnames])
    strata <- as.factor(strata)

    # Pushing down checks to prop_strat_wilson
  } else if (checkmate::test_subset(method, c("strat_wilson", "strat_wilsonc"))) {
    stop("To use stratified methods you need to specify the strata variables.")
  }
  if (checkmate::test_atomic_vector(df)) {
    rsp <- as.logical(df)
  } else {
    rsp <- as.logical(df[[.var]])
  }
  n <- sum(rsp)
  p_hat <- mean(rsp)

  prop_ci <- switch(method,
    "clopper-pearson" = prop_clopper_pearson(rsp, conf_level),
    "wilson" = prop_wilson(rsp, conf_level),
    "wilsonc" = prop_wilson(rsp, conf_level, correct = TRUE),
    "strat_wilson" = prop_strat_wilson(rsp,
      strata,
      weights,
      conf_level,
      max_iterations,
      correct = FALSE
    )$conf_int,
    "strat_wilsonc" = prop_strat_wilson(rsp,
      strata,
      weights,
      conf_level,
      max_iterations,
      correct = TRUE
    )$conf_int,
    "wald" = prop_wald(rsp, conf_level),
    "waldcc" = prop_wald(rsp, conf_level, correct = TRUE),
    "agresti-coull" = prop_agresti_coull(rsp, conf_level),
    "jeffreys" = prop_jeffreys(rsp, conf_level)
  )

  list(
    "n_prop" = formatters::with_label(c(n, p_hat), "Responders"),
    "prop_ci" = formatters::with_label(
      x = 100 * prop_ci, label = d_proportion(conf_level, method, long = long)
    )
  )
}

#' @describeIn estimate_proportion Formatted analysis function which is used as `afun`
#'   in `estimate_proportion()`.
#'
#' @return
#' * `a_proportion()` returns the corresponding list with formatted [rtables::CellValue()].
#'
#' @export
a_proportion <- make_afun(
  s_proportion,
  .formats = c(n_prop = "xx (xx.x%)", prop_ci = "(xx.x, xx.x)")
)

#' @describeIn estimate_proportion Layout-creating function which can take statistics function arguments
#'   and additional format arguments. This function is a wrapper for [rtables::analyze()].
#'
#' @return
#' * `estimate_proportion()` returns a layout object suitable for passing to further layouting functions,
#'   or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
#'   the statistics from `s_proportion()` to the table layout.
#'
#' @examples
#' dta_test <- data.frame(
#'   USUBJID = paste0("S", 1:12),
#'   ARM     = rep(LETTERS[1:3], each = 4),
#'   AVAL    = rep(LETTERS[1:3], each = 4)
#' )
#'
#' basic_table() %>%
#'   split_cols_by("ARM") %>%
#'   estimate_proportion(vars = "AVAL") %>%
#'   build_table(df = dta_test)
#'
#' @export
#' @order 2
estimate_proportion <- function(lyt,
                                vars,
                                conf_level = 0.95,
                                method = c(
                                  "waldcc", "wald", "clopper-pearson",
                                  "wilson", "wilsonc", "strat_wilson", "strat_wilsonc",
                                  "agresti-coull", "jeffreys"
                                ),
                                weights = NULL,
                                max_iterations = 50,
                                variables = list(strata = NULL),
                                long = FALSE,
                                na_str = default_na_str(),
                                nested = TRUE,
                                ...,
                                show_labels = "hidden",
                                table_names = vars,
                                .stats = NULL,
                                .formats = NULL,
                                .labels = NULL,
                                .indent_mods = NULL) {
  extra_args <- list(
    conf_level = conf_level, method = method, weights = weights, max_iterations = max_iterations,
    variables = variables, long = long, ...
  )

  afun <- make_afun(
    a_proportion,
    .stats = .stats,
    .formats = .formats,
    .labels = .labels,
    .indent_mods = .indent_mods
  )
  analyze(
    lyt,
    vars,
    afun = afun,
    na_str = na_str,
    nested = nested,
    extra_args = extra_args,
    show_labels = show_labels,
    table_names = table_names
  )
}

#' Helper functions for calculating proportion confidence intervals
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Functions to calculate different proportion confidence intervals for use in [estimate_proportion()].
#'
#' @inheritParams argument_convention
#' @inheritParams estimate_proportion
#'
#' @return Confidence interval of a proportion.
#'
#' @seealso [estimate_proportion], descriptive function [d_proportion()],
#'  and helper functions [strata_normal_quantile()] and [update_weights_strat_wilson()].
#'
#' @name h_proportions
NULL

#' @describeIn h_proportions Calculates the Wilson interval by calling [stats::prop.test()].
#'  Also referred to as Wilson score interval.
#'
#' @examples
#' rsp <- c(
#'   TRUE, TRUE, TRUE, TRUE, TRUE,
#'   FALSE, FALSE, FALSE, FALSE, FALSE
#' )
#' prop_wilson(rsp, conf_level = 0.9)
#'
#' @export
prop_wilson <- function(rsp, conf_level, correct = FALSE) {
  y <- stats::prop.test(
    sum(rsp),
    length(rsp),
    correct = correct,
    conf.level = conf_level
  )

  as.numeric(y$conf.int)
}

#' @describeIn h_proportions Calculates the stratified Wilson confidence
#'   interval for unequal proportions as described in \insertCite{Yan2010-jt;textual}{tern}
#'
#' @param strata (`factor`)\cr variable with one level per stratum and same length as `rsp`.
#' @param weights (`numeric` or `NULL`)\cr weights for each level of the strata. If `NULL`, they are
#'   estimated using the iterative algorithm proposed in \insertCite{Yan2010-jt;textual}{tern} that
#'   minimizes the weighted squared length of the confidence interval.
#' @param max_iterations (`count`)\cr maximum number of iterations for the iterative procedure used
#'   to find estimates of optimal weights.
#' @param correct (`flag`)\cr whether to include the continuity correction. For further information, see for example
#'   for [stats::prop.test()].
#'
#' @references
#' \insertRef{Yan2010-jt}{tern}
#'
#' @examples
#' # Stratified Wilson confidence interval with unequal probabilities
#'
#' set.seed(1)
#' rsp <- sample(c(TRUE, FALSE), 100, TRUE)
#' strata_data <- data.frame(
#'   "f1" = sample(c("a", "b"), 100, TRUE),
#'   "f2" = sample(c("x", "y", "z"), 100, TRUE),
#'   stringsAsFactors = TRUE
#' )
#' strata <- interaction(strata_data)
#' n_strata <- ncol(table(rsp, strata)) # Number of strata
#'
#' prop_strat_wilson(
#'   rsp = rsp, strata = strata,
#'   conf_level = 0.90
#' )
#'
#' # Not automatic setting of weights
#' prop_strat_wilson(
#'   rsp = rsp, strata = strata,
#'   weights = rep(1 / n_strata, n_strata),
#'   conf_level = 0.90
#' )
#'
#' @export
prop_strat_wilson <- function(rsp,
                              strata,
                              weights = NULL,
                              conf_level = 0.95,
                              max_iterations = NULL,
                              correct = FALSE) {
  checkmate::assert_logical(rsp, any.missing = FALSE)
  checkmate::assert_factor(strata, len = length(rsp))
  assert_proportion_value(conf_level)

  tbl <- table(rsp, strata)
  n_strata <- length(unique(strata))

  # Checking the weights and maximum number of iterations.
  do_iter <- FALSE
  if (is.null(weights)) {
    weights <- rep(1 / n_strata, n_strata) # Initialization for iterative procedure
    do_iter <- TRUE

    # Iteration parameters
    if (is.null(max_iterations)) max_iterations <- 10
    checkmate::assert_int(max_iterations, na.ok = FALSE, null.ok = FALSE, lower = 1)
  }
  checkmate::assert_numeric(weights, lower = 0, upper = 1, any.missing = FALSE, len = n_strata)
  sum_weights <- checkmate::assert_int(sum(weights))
  if (as.integer(sum_weights + 0.5) != 1L) stop("Sum of weights must be 1L.")

  xs <- tbl["TRUE", ]
  ns <- colSums(tbl)
  use_stratum <- (ns > 0)
  ns <- ns[use_stratum]
  xs <- xs[use_stratum]
  ests <- xs / ns
  vars <- ests * (1 - ests) / ns

  strata_qnorm <- strata_normal_quantile(vars, weights, conf_level)

  # Iterative setting of weights if they were not set externally
  weights_new <- if (do_iter) {
    update_weights_strat_wilson(vars, strata_qnorm, weights, ns, max_iterations, conf_level)$weights
  } else {
    weights
  }

  strata_conf_level <- 2 * stats::pnorm(strata_qnorm) - 1

  ci_by_strata <- Map(
    function(x, n) {
      # Classic Wilson's confidence interval
      suppressWarnings(stats::prop.test(x, n, correct = correct, conf.level = strata_conf_level)$conf.int)
    },
    x = xs,
    n = ns
  )
  lower_by_strata <- sapply(ci_by_strata, "[", 1L)
  upper_by_strata <- sapply(ci_by_strata, "[", 2L)

  lower <- sum(weights_new * lower_by_strata)
  upper <- sum(weights_new * upper_by_strata)

  # Return values
  if (do_iter) {
    list(
      conf_int = c(
        lower = lower,
        upper = upper
      ),
      weights = weights_new
    )
  } else {
    list(
      conf_int = c(
        lower = lower,
        upper = upper
      )
    )
  }
}

#' @describeIn h_proportions Calculates the Clopper-Pearson interval by calling [stats::binom.test()].
#'   Also referred to as the `exact` method.
#'
#' @examples
#' prop_clopper_pearson(rsp, conf_level = .95)
#'
#' @export
prop_clopper_pearson <- function(rsp,
                                 conf_level) {
  y <- stats::binom.test(
    x = sum(rsp),
    n = length(rsp),
    conf.level = conf_level
  )
  as.numeric(y$conf.int)
}

#' @describeIn h_proportions Calculates the Wald interval by following the usual textbook definition
#'   for a single proportion confidence interval using the normal approximation.
#'
#' @param correct (`flag`)\cr whether to apply continuity correction.
#'
#' @examples
#' prop_wald(rsp, conf_level = 0.95)
#' prop_wald(rsp, conf_level = 0.95, correct = TRUE)
#'
#' @export
prop_wald <- function(rsp, conf_level, correct = FALSE) {
  n <- length(rsp)
  p_hat <- mean(rsp)
  z <- stats::qnorm((1 + conf_level) / 2)
  q_hat <- 1 - p_hat
  correct <- if (correct) 1 / (2 * n) else 0

  err <- z * sqrt(p_hat * q_hat) / sqrt(n) + correct
  l_ci <- max(0, p_hat - err)
  u_ci <- min(1, p_hat + err)

  c(l_ci, u_ci)
}

#' @describeIn h_proportions Calculates the Agresti-Coull interval. Constructed (for 95% CI) by adding two successes
#'   and two failures to the data and then using the Wald formula to construct a CI.
#'
#' @examples
#' prop_agresti_coull(rsp, conf_level = 0.95)
#'
#' @export
prop_agresti_coull <- function(rsp, conf_level) {
  n <- length(rsp)
  x_sum <- sum(rsp)
  z <- stats::qnorm((1 + conf_level) / 2)

  # Add here both z^2 / 2 successes and failures.
  x_sum_tilde <- x_sum + z^2 / 2
  n_tilde <- n + z^2

  # Then proceed as with the Wald interval.
  p_tilde <- x_sum_tilde / n_tilde
  q_tilde <- 1 - p_tilde
  err <- z * sqrt(p_tilde * q_tilde) / sqrt(n_tilde)
  l_ci <- max(0, p_tilde - err)
  u_ci <- min(1, p_tilde + err)

  c(l_ci, u_ci)
}

#' @describeIn h_proportions Calculates the Jeffreys interval, an equal-tailed interval based on the
#'   non-informative Jeffreys prior for a binomial proportion.
#'
#' @examples
#' prop_jeffreys(rsp, conf_level = 0.95)
#'
#' @export
prop_jeffreys <- function(rsp,
                          conf_level) {
  n <- length(rsp)
  x_sum <- sum(rsp)

  alpha <- 1 - conf_level
  l_ci <- ifelse(
    x_sum == 0,
    0,
    stats::qbeta(alpha / 2, x_sum + 0.5, n - x_sum + 0.5)
  )

  u_ci <- ifelse(
    x_sum == n,
    1,
    stats::qbeta(1 - alpha / 2, x_sum + 0.5, n - x_sum + 0.5)
  )

  c(l_ci, u_ci)
}

#' Description of the proportion summary
#'
#' @description `r lifecycle::badge("stable")`
#'
#' This is a helper function that describes the analysis in [s_proportion()].
#'
#' @inheritParams s_proportion
#' @param long (`flag`)\cr whether a long or a short (default) description is required.
#'
#' @return String describing the analysis.
#'
#' @export
d_proportion <- function(conf_level,
                         method,
                         long = FALSE) {
  label <- paste0(conf_level * 100, "% CI")

  if (long) label <- paste(label, "for Response Rates")

  method_part <- switch(method,
    "clopper-pearson" = "Clopper-Pearson",
    "waldcc" = "Wald, with correction",
    "wald" = "Wald, without correction",
    "wilson" = "Wilson, without correction",
    "strat_wilson" = "Stratified Wilson, without correction",
    "wilsonc" = "Wilson, with correction",
    "strat_wilsonc" = "Stratified Wilson, with correction",
    "agresti-coull" = "Agresti-Coull",
    "jeffreys" = "Jeffreys",
    stop(paste(method, "does not have a description"))
  )

  paste0(label, " (", method_part, ")")
}

#' Helper function for the estimation of stratified quantiles
#'
#' @description `r lifecycle::badge("stable")`
#'
#' This function wraps the estimation of stratified percentiles when we assume
#' the approximation for large numbers. This is necessary only in the case
#' proportions for each strata are unequal.
#'
#' @inheritParams argument_convention
#' @inheritParams prop_strat_wilson
#'
#' @return Stratified quantile.
#'
#' @seealso [prop_strat_wilson()]
#'
#' @examples
#' strata_data <- table(data.frame(
#'   "f1" = sample(c(TRUE, FALSE), 100, TRUE),
#'   "f2" = sample(c("x", "y", "z"), 100, TRUE),
#'   stringsAsFactors = TRUE
#' ))
#' ns <- colSums(strata_data)
#' ests <- strata_data["TRUE", ] / ns
#' vars <- ests * (1 - ests) / ns
#' weights <- rep(1 / length(ns), length(ns))
#'
#' strata_normal_quantile(vars, weights, 0.95)
#'
#' @export
strata_normal_quantile <- function(vars, weights, conf_level) {
  summands <- weights^2 * vars
  # Stratified quantile
  sqrt(sum(summands)) / sum(sqrt(summands)) * stats::qnorm((1 + conf_level) / 2)
}

#' Helper function for the estimation of weights for `prop_strat_wilson()`
#'
#' @description `r lifecycle::badge("stable")`
#'
#' This function wraps the iteration procedure that allows you to estimate
#' the weights for each proportional strata. This assumes to minimize the
#' weighted squared length of the confidence interval.
#'
#' @inheritParams prop_strat_wilson
#' @param vars (`numeric`)\cr normalized proportions for each strata.
#' @param strata_qnorm (`numeric(1)`)\cr initial estimation with identical weights of the quantiles.
#' @param initial_weights (`numeric`)\cr initial weights used to calculate `strata_qnorm`. This can
#'   be optimized in the future if we need to estimate better initial weights.
#' @param n_per_strata (`numeric`)\cr number of elements in each strata.
#' @param max_iterations (`integer(1)`)\cr maximum number of iterations to be tried. Convergence is always checked.
#' @param tol (`numeric(1)`)\cr tolerance threshold for convergence.
#'
#' @return A `list` of 3 elements: `n_it`, `weights`, and `diff_v`.
#'
#' @seealso For references and details see [prop_strat_wilson()].
#'
#' @examples
#' vs <- c(0.011, 0.013, 0.012, 0.014, 0.017, 0.018)
#' sq <- 0.674
#' ws <- rep(1 / length(vs), length(vs))
#' ns <- c(22, 18, 17, 17, 14, 12)
#'
#' update_weights_strat_wilson(vs, sq, ws, ns, 100, 0.95, 0.001)
#'
#' @export
update_weights_strat_wilson <- function(vars,
                                        strata_qnorm,
                                        initial_weights,
                                        n_per_strata,
                                        max_iterations = 50,
                                        conf_level = 0.95,
                                        tol = 0.001) {
  it <- 0
  diff_v <- NULL

  while (it < max_iterations) {
    it <- it + 1
    weights_new_t <- (1 + strata_qnorm^2 / n_per_strata)^2
    weights_new_b <- (vars + strata_qnorm^2 / (4 * n_per_strata^2))
    weights_new <- weights_new_t / weights_new_b
    weights_new <- weights_new / sum(weights_new)
    strata_qnorm <- strata_normal_quantile(vars, weights_new, conf_level)
    diff_v <- c(diff_v, sum(abs(weights_new - initial_weights)))
    if (diff_v[length(diff_v)] < tol) break
    initial_weights <- weights_new
  }

  if (it == max_iterations) {
    warning("The heuristic to find weights did not converge with max_iterations = ", max_iterations)
  }

  list(
    "n_it" = it,
    "weights" = weights_new,
    "diff_v" = diff_v
  )
}

# Utility functions to cooperate with {rtables} package

#' Convert table into matrix of strings
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Helper function to use mostly within tests. `with_spaces`parameter allows
#' to test not only for content but also indentation and table structure.
#' `print_txt_to_copy` instead facilitate the testing development by returning a well
#' formatted text that needs only to be copied and pasted in the expected output.
#'
#' @inheritParams formatters::toString
#' @param x (`VTableTree`)\cr `rtables` table object.
#' @param with_spaces (`flag`)\cr whether the tested table should keep the indentation and other relevant spaces.
#' @param print_txt_to_copy (`flag`)\cr utility to have a way to copy the input table directly
#'   into the expected variable instead of copying it too manually.
#'
#' @return A `matrix` of `string`s. If `print_txt_to_copy = TRUE` the well formatted printout of the
#'   table will be printed to console, ready to be copied as a expected value.
#'
#' @examples
#' tbl <- basic_table() %>%
#'   split_rows_by("SEX") %>%
#'   split_cols_by("ARM") %>%
#'   analyze("AGE") %>%
#'   build_table(tern_ex_adsl)
#'
#' to_string_matrix(tbl, widths = ceiling(propose_column_widths(tbl) / 2))
#'
#' @export
to_string_matrix <- function(x, widths = NULL, max_width = NULL,
                             hsep = formatters::default_hsep(),
                             with_spaces = TRUE, print_txt_to_copy = FALSE) {
  checkmate::assert_flag(with_spaces)
  checkmate::assert_flag(print_txt_to_copy)
  checkmate::assert_int(max_width, null.ok = TRUE)

  if (inherits(x, "MatrixPrintForm")) {
    tx <- x
  } else {
    tx <- matrix_form(x, TRUE)
  }

  tf_wrap <- FALSE
  if (!is.null(max_width)) {
    tf_wrap <- TRUE
  }

  # Producing the matrix to test
  if (with_spaces) {
    out <- strsplit(toString(tx, widths = widths, tf_wrap = tf_wrap, max_width = max_width, hsep = hsep), "\n")[[1]]
  } else {
    out <- tx$strings
  }

  # Printing to console formatted output that needs to be copied in "expected"
  if (print_txt_to_copy) {
    out_tmp <- out
    if (!with_spaces) {
      out_tmp <- apply(out, 1, paste0, collapse = '", "')
    }
    cat(paste0('c(\n  "', paste0(out_tmp, collapse = '",\n  "'), '"\n)'))
  }

  # Return values
  return(out)
}

#' Blank for missing input
#'
#' Helper function to use in tabulating model results.
#'
#' @param x (`vector`)\cr input for a cell.
#'
#' @return An empty `character` vector if all entries in `x` are missing (`NA`), otherwise
#'   the unlisted version of `x`.
#'
#' @keywords internal
unlist_and_blank_na <- function(x) {
  unl <- unlist(x)
  if (all(is.na(unl))) {
    character()
  } else {
    unl
  }
}

#' Constructor for content functions given a data frame with flag input
#'
#' This can be useful for tabulating model results.
#'
#' @param analysis_var (`string`)\cr variable name for the column containing values to be returned by the
#'   content function.
#' @param flag_var (`string`)\cr variable name for the logical column identifying which row should be returned.
#' @param format (`string`)\cr `rtables` format to use.
#'
#' @return A content function which gives `df$analysis_var` at the row identified by
#'   `.df_row$flag` in the given format.
#'
#' @keywords internal
cfun_by_flag <- function(analysis_var,
                         flag_var,
                         format = "xx",
                         .indent_mods = NULL) {
  checkmate::assert_string(analysis_var)
  checkmate::assert_string(flag_var)
  function(df, labelstr) {
    row_index <- which(df[[flag_var]])
    x <- unlist_and_blank_na(df[[analysis_var]][row_index])
    formatters::with_label(
      rcell(x, format = format, indent_mod = .indent_mods),
      labelstr
    )
  }
}

#' Content row function to add row total to labels
#'
#' This takes the label of the latest row split level and adds the row total from `df` in parentheses.
#' This function differs from [c_label_n_alt()] by taking row counts from `df` rather than
#' `alt_counts_df`, and is used by [add_rowcounts()] when `alt_counts` is set to `FALSE`.
#'
#' @inheritParams argument_convention
#'
#' @return A list with formatted [rtables::CellValue()] with the row count value and the correct label.
#'
#' @note It is important here to not use `df` but rather `.N_row` in the implementation, because
#'   the former is already split by columns and will refer to the first column of the data only.
#'
#' @seealso [c_label_n_alt()] which performs the same function but retrieves row counts from
#'   `alt_counts_df` instead of `df`.
#'
#' @keywords internal
c_label_n <- function(df,
                      labelstr,
                      .N_row) { # nolint
  label <- paste0(labelstr, " (N=", .N_row, ")")
  in_rows(
    .list = list(row_count = formatters::with_label(c(.N_row, .N_row), label)),
    .formats = c(row_count = function(x, ...) "")
  )
}

#' Content row function to add `alt_counts_df` row total to labels
#'
#' This takes the label of the latest row split level and adds the row total from `alt_counts_df`
#' in parentheses. This function differs from [c_label_n()] by taking row counts from `alt_counts_df`
#' rather than `df`, and is used by [add_rowcounts()] when `alt_counts` is set to `TRUE`.
#'
#' @inheritParams argument_convention
#'
#' @return A list with formatted [rtables::CellValue()] with the row count value and the correct label.
#'
#' @seealso [c_label_n()] which performs the same function but retrieves row counts from `df` instead
#'   of `alt_counts_df`.
#'
#' @keywords internal
c_label_n_alt <- function(df,
                          labelstr,
                          .alt_df_row) {
  N_row_alt <- nrow(.alt_df_row) # nolint
  label <- paste0(labelstr, " (N=", N_row_alt, ")")
  in_rows(
    .list = list(row_count = formatters::with_label(c(N_row_alt, N_row_alt), label)),
    .formats = c(row_count = function(x, ...) "")
  )
}

#' Layout-creating function to add row total counts
#'
#' @description `r lifecycle::badge("stable")`
#'
#' This works analogously to [rtables::add_colcounts()] but on the rows. This function
#'  is a wrapper for [rtables::summarize_row_groups()].
#'
#' @inheritParams argument_convention
#' @param alt_counts (`flag`)\cr whether row counts should be taken from `alt_counts_df` (`TRUE`)
#'   or from `df` (`FALSE`). Defaults to `FALSE`.
#'
#' @return A modified layout where the latest row split labels now have the row-wise
#'   total counts (i.e. without column-based subsetting) attached in parentheses.
#'
#' @note Row count values are contained in these row count rows but are not displayed
#'   so that they are not considered zero rows by default when pruning.
#'
#' @examples
#' basic_table() %>%
#'   split_cols_by("ARM") %>%
#'   add_colcounts() %>%
#'   split_rows_by("RACE", split_fun = drop_split_levels) %>%
#'   add_rowcounts() %>%
#'   analyze("AGE", afun = list_wrap_x(summary), format = "xx.xx") %>%
#'   build_table(DM)
#'
#' @export
add_rowcounts <- function(lyt, alt_counts = FALSE) {
  summarize_row_groups(
    lyt,
    cfun = if (alt_counts) c_label_n_alt else c_label_n
  )
}

#' Obtain column indices
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Helper function to extract column indices from a `VTableTree` for a given
#' vector of column names.
#'
#' @param table_tree (`VTableTree`)\cr `rtables` table object to extract the indices from.
#' @param col_names (`character`)\cr vector of column names.
#'
#' @return A vector of column indices.
#'
#' @export
h_col_indices <- function(table_tree, col_names) {
  checkmate::assert_class(table_tree, "VTableNodeInfo")
  checkmate::assert_subset(col_names, names(attr(col_info(table_tree), "cextra_args")), empty.ok = FALSE)
  match(col_names, names(attr(col_info(table_tree), "cextra_args")))
}

#' Labels or names of list elements
#'
#' Internal helper function for working with nested statistic function results which typically
#' don't have labels but names that we can use.
#'
#' @param x (`list`)\cr a list.
#'
#' @return A `character` vector with the labels or names for the list elements.
#'
#' @keywords internal
labels_or_names <- function(x) {
  checkmate::assert_multi_class(x, c("data.frame", "list"))
  labs <- sapply(x, obj_label)
  nams <- rlang::names2(x)
  label_is_null <- sapply(labs, is.null)
  result <- unlist(ifelse(label_is_null, nams, labs))
  return(result)
}

#' Convert to `rtable`
#'
#' @description `r lifecycle::badge("stable")`
#'
#' This is a new generic function to convert objects to `rtable` tables.
#'
#' @param x (`data.frame`)\cr the object which should be converted to an `rtable`.
#' @param ... additional arguments for methods.
#'
#' @return An `rtables` table object. Note that the concrete class will depend on the method used.
#'
#' @export
as.rtable <- function(x, ...) { # nolint
  UseMethod("as.rtable", x)
}

#' @describeIn as.rtable Method for converting a `data.frame` that contains numeric columns to `rtable`.
#'
#' @param format (`string` or `function`)\cr the format which should be used for the columns.
#'
#' @method as.rtable data.frame
#'
#' @examples
#' x <- data.frame(
#'   a = 1:10,
#'   b = rnorm(10)
#' )
#' as.rtable(x)
#'
#' @export
as.rtable.data.frame <- function(x, format = "xx.xx", ...) {
  checkmate::assert_numeric(unlist(x))
  do.call(
    rtable,
    c(
      list(
        header = labels_or_names(x),
        format = format
      ),
      Map(
        function(row, row_name) {
          do.call(
            rrow,
            c(as.list(unname(row)),
              row.name = row_name
            )
          )
        },
        row = as.data.frame(t(x)),
        row_name = rownames(x)
      )
    )
  )
}

#' Split parameters
#'
#' @description `r lifecycle::badge("stable")`
#'
#' It divides the data in the vector `param` into the groups defined by `f` based on specified `values`. It is relevant
#' in `rtables` layers so as to distribute parameters `.stats` or' `.formats` into lists with items corresponding to
#' specific analysis function.
#'
#' @param param (`vector`)\cr the parameter to be split.
#' @param value (`vector`)\cr the value used to split.
#' @param f (`list`)\cr the reference to make the split.
#'
#' @return A named `list` with the same element names as `f`, each containing the elements specified in `.stats`.
#'
#' @examples
#' f <- list(
#'   surv = c("pt_at_risk", "event_free_rate", "rate_se", "rate_ci"),
#'   surv_diff = c("rate_diff", "rate_diff_ci", "ztest_pval")
#' )
#'
#' .stats <- c("pt_at_risk", "rate_diff")
#' h_split_param(.stats, .stats, f = f)
#'
#' # $surv
#' # [1] "pt_at_risk"
#' #
#' # $surv_diff
#' # [1] "rate_diff"
#'
#' .formats <- c("pt_at_risk" = "xx", "event_free_rate" = "xxx")
#' h_split_param(.formats, names(.formats), f = f)
#'
#' # $surv
#' # pt_at_risk event_free_rate
#' # "xx"           "xxx"
#' #
#' # $surv_diff
#' # NULL
#'
#' @export
h_split_param <- function(param,
                          value,
                          f) {
  y <- lapply(f, function(x) param[value %in% x])
  lapply(y, function(x) if (length(x) == 0) NULL else x)
}

#' Get selected statistics names
#'
#' Helper function to be used for creating `afun`.
#'
#' @param .stats (`vector` or `NULL`)\cr input to the layout creating function. Note that `NULL` means
#'   in this context that all default statistics should be used.
#' @param all_stats (`character`)\cr all statistics which can be selected here potentially.
#'
#' @return A `character` vector with the selected statistics.
#'
#' @keywords internal
afun_selected_stats <- function(.stats, all_stats) {
  checkmate::assert_character(.stats, null.ok = TRUE)
  checkmate::assert_character(all_stats)
  if (is.null(.stats)) {
    all_stats
  } else {
    intersect(.stats, all_stats)
  }
}

#' Add variable labels to top left corner in table
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Helper layout-creating function to append the variable labels of a given variables vector
#' from a given dataset in the top left corner. If a variable label is not found then the
#' variable name itself is used instead. Multiple variable labels are concatenated with slashes.
#'
#' @inheritParams argument_convention
#' @param vars (`character`)\cr variable names of which the labels are to be looked up in `df`.
#' @param indent (`integer(1)`)\cr non-negative number of nested indent space, default to 0L which means no indent.
#'   1L means two spaces indent, 2L means four spaces indent and so on.
#'
#' @return A modified layout with the new variable label(s) added to the top-left material.
#'
#' @note This is not an optimal implementation of course, since we are using here the data set
#'   itself during the layout creation. When we have a more mature `rtables` implementation then
#'   this will also be improved or not necessary anymore.
#'
#' @examples
#' lyt <- basic_table() %>%
#'   split_cols_by("ARM") %>%
#'   add_colcounts() %>%
#'   split_rows_by("SEX") %>%
#'   append_varlabels(DM, "SEX") %>%
#'   analyze("AGE", afun = mean) %>%
#'   append_varlabels(DM, "AGE", indent = 1)
#' build_table(lyt, DM)
#'
#' lyt <- basic_table() %>%
#'   split_cols_by("ARM") %>%
#'   split_rows_by("SEX") %>%
#'   analyze("AGE", afun = mean) %>%
#'   append_varlabels(DM, c("SEX", "AGE"))
#' build_table(lyt, DM)
#'
#' @export
append_varlabels <- function(lyt, df, vars, indent = 0L) {
  if (checkmate::test_flag(indent)) {
    warning("indent argument is now accepting integers. Boolean indent will be converted to integers.")
    indent <- as.integer(indent)
  }

  checkmate::assert_data_frame(df)
  checkmate::assert_character(vars)
  checkmate::assert_count(indent)

  lab <- formatters::var_labels(df[vars], fill = TRUE)
  lab <- paste(lab, collapse = " / ")
  space <- paste(rep(" ", indent * 2), collapse = "")
  lab <- paste0(space, lab)

  append_topleft(lyt, lab)
}

#' Default string replacement for `NA` values
#'
#' @description `r lifecycle::badge("stable")`
#'
#' The default string used to represent `NA` values. This value is used as the default
#' value for the `na_str` argument throughout the `tern` package, and printed in place
#' of `NA` values in output tables. If not specified for each `tern` function by the user
#' via the `na_str` argument, or in the R environment options via [set_default_na_str()],
#' then `NA` is used.
#'
#' @param na_str (`string`)\cr single string value to set in the R environment options as
#'   the default value to replace `NA`s. Use `getOption("tern_default_na_str")` to check the
#'   current value set in the R environment (defaults to `NULL` if not set).
#'
#' @name default_na_str
NULL

#' @describeIn default_na_str Accessor for default `NA` value replacement string.
#'
#' @return
#' * `default_na_str` returns the current value if an R environment option has been set
#'   for `"tern_default_na_str"`, or `NA_character_` otherwise.
#'
#' @examples
#' # Default settings
#' default_na_str()
#' getOption("tern_default_na_str")
#'
#' # Set custom value
#' set_default_na_str("<Missing>")
#'
#' # Settings after value has been set
#' default_na_str()
#' getOption("tern_default_na_str")
#'
#' @export
default_na_str <- function() {
  getOption("tern_default_na_str", default = NA_character_)
}

#' @describeIn default_na_str Setter for default `NA` value replacement string. Sets the
#'   option `"tern_default_na_str"` within the R environment.
#'
#' @return
#' * `set_default_na_str` has no return value.
#'
#' @export
set_default_na_str <- function(na_str) {
  checkmate::assert_character(na_str, len = 1, null.ok = TRUE)
  options("tern_default_na_str" = na_str)
}

#' Helper functions for subgroup treatment effect pattern (STEP) calculations
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Helper functions that are used internally for the STEP calculations.
#'
#' @inheritParams argument_convention
#'
#' @name h_step
#' @include control_step.R
NULL

#' @describeIn h_step Creates the windows for STEP, based on the control settings
#'   provided.
#'
#' @param x (`numeric`)\cr biomarker value(s) to use (without `NA`).
#' @param control (named `list`)\cr output from `control_step()`.
#'
#' @return
#' * `h_step_window()` returns a list containing the window-selection matrix `sel`
#'   and the interval information matrix `interval`.
#'
#' @export
h_step_window <- function(x,
                          control = control_step()) {
  checkmate::assert_numeric(x, min.len = 1, any.missing = FALSE)
  checkmate::assert_list(control, names = "named")

  sel <- matrix(FALSE, length(x), control$num_points)
  out <- matrix(0, control$num_points, 3)
  colnames(out) <- paste("Interval", c("Center", "Lower", "Upper"))
  if (control$use_percentile) {
    # Create windows according to percentile cutoffs.
    out <- cbind(out, out)
    colnames(out)[1:3] <- paste("Percentile", c("Center", "Lower", "Upper"))
    xs <- seq(0, 1, length.out = control$num_points + 2)[-1]
    for (i in seq_len(control$num_points)) {
      out[i, 2:3] <- c(
        max(xs[i] - control$bandwidth, 0),
        min(xs[i] + control$bandwidth, 1)
      )
      out[i, 5:6] <- stats::quantile(x, out[i, 2:3])
      sel[, i] <- x >= out[i, 5] & x <= out[i, 6]
    }
    # Center is the middle point of the percentile window.
    out[, 1] <- xs[-control$num_points - 1]
    out[, 4] <- stats::quantile(x, out[, 1])
  } else {
    # Create windows according to cutoffs.
    m <- c(min(x), max(x))
    xs <- seq(m[1], m[2], length.out = control$num_points + 2)[-1]
    for (i in seq_len(control$num_points)) {
      out[i, 2:3] <- c(
        max(xs[i] - control$bandwidth, m[1]),
        min(xs[i] + control$bandwidth, m[2])
      )
      sel[, i] <- x >= out[i, 2] & x <= out[i, 3]
    }
    # Center is the same as the point for predicting.
    out[, 1] <- xs[-control$num_points - 1]
  }
  list(sel = sel, interval = out)
}

#' @describeIn h_step Calculates the estimated treatment effect estimate
#'   on the linear predictor scale and corresponding standard error from a STEP `model` fitted
#'   on `data` given `variables` specification, for a single biomarker value `x`.
#'   This works for both `coxph` and `glm` models, i.e. for calculating log hazard ratio or log odds
#'   ratio estimates.
#'
#' @param model (`coxph` or `glm`)\cr the regression model object.
#'
#' @return
#' * `h_step_trt_effect()` returns a vector with elements `est` and `se`.
#'
#' @export
h_step_trt_effect <- function(data,
                              model,
                              variables,
                              x) {
  checkmate::assert_multi_class(model, c("coxph", "glm"))
  checkmate::assert_number(x)
  assert_df_with_variables(data, variables)
  checkmate::assert_factor(data[[variables$arm]], n.levels = 2)

  newdata <- data[c(1, 1), ]
  newdata[, variables$biomarker] <- x
  newdata[, variables$arm] <- levels(data[[variables$arm]])
  model_terms <- stats::delete.response(stats::terms(model))
  model_frame <- stats::model.frame(model_terms, data = newdata, xlev = model$xlevels)
  mat <- stats::model.matrix(model_terms, data = model_frame, contrasts.arg = model$contrasts)
  coefs <- stats::coef(model)
  # Note: It is important to use the coef subset from matrix, otherwise intercept and
  # strata are included for coxph() models.
  mat <- mat[, names(coefs)]
  mat_diff <- diff(mat)
  est <- mat_diff %*% coefs
  var <- mat_diff %*% stats::vcov(model) %*% t(mat_diff)
  se <- sqrt(var)
  c(
    est = est,
    se = se
  )
}

#' @describeIn h_step Builds the model formula used in survival STEP calculations.
#'
#' @return
#' * `h_step_survival_formula()` returns a model formula.
#'
#' @export
h_step_survival_formula <- function(variables,
                                    control = control_step()) {
  checkmate::assert_character(variables$covariates, null.ok = TRUE)

  assert_list_of_variables(variables[c("arm", "biomarker", "event", "time")])
  form <- paste0("Surv(", variables$time, ", ", variables$event, ") ~ ", variables$arm)
  if (control$degree > 0) {
    form <- paste0(form, " * stats::poly(", variables$biomarker, ", degree = ", control$degree, ", raw = TRUE)")
  }
  if (!is.null(variables$covariates)) {
    form <- paste(form, "+", paste(variables$covariates, collapse = "+"))
  }
  if (!is.null(variables$strata)) {
    form <- paste0(form, " + strata(", paste0(variables$strata, collapse = ", "), ")")
  }
  stats::as.formula(form)
}

#' @describeIn h_step Estimates the model with `formula` built based on
#'   `variables` in `data` for a given `subset` and `control` parameters for the
#'   Cox regression.
#'
#' @param formula (`formula`)\cr the regression model formula.
#' @param subset (`logical`)\cr subset vector.
#'
#' @return
#' * `h_step_survival_est()` returns a matrix of number of observations `n`,
#'   `events`, log hazard ratio estimates `loghr`, standard error `se`,
#'   and Wald confidence interval bounds `ci_lower` and `ci_upper`. One row is
#'   included for each biomarker value in `x`.
#'
#' @export
h_step_survival_est <- function(formula,
                                data,
                                variables,
                                x,
                                subset = rep(TRUE, nrow(data)),
                                control = control_coxph()) {
  checkmate::assert_formula(formula)
  assert_df_with_variables(data, variables)
  checkmate::assert_logical(subset, min.len = 1, any.missing = FALSE)
  checkmate::assert_numeric(x, min.len = 1, any.missing = FALSE)
  checkmate::assert_list(control, names = "named")

  # Note: `subset` in `coxph` needs to be an expression referring to `data` variables.
  data$.subset <- subset
  coxph_warnings <- NULL
  tryCatch(
    withCallingHandlers(
      expr = {
        fit <- survival::coxph(
          formula = formula,
          data = data,
          subset = .subset,
          ties = control$ties
        )
      },
      warning = function(w) {
        coxph_warnings <<- c(coxph_warnings, w)
        invokeRestart("muffleWarning")
      }
    ),
    finally = {
    }
  )
  if (!is.null(coxph_warnings)) {
    warning(paste(
      "Fit warnings occurred, please consider using a simpler model, or",
      "larger `bandwidth`, less `num_points` in `control_step()` settings"
    ))
  }
  # Produce a matrix with one row per `x` and columns `est` and `se`.
  estimates <- t(vapply(
    X = x,
    FUN = h_step_trt_effect,
    FUN.VALUE = c(1, 2),
    data = data,
    model = fit,
    variables = variables
  ))
  q_norm <- stats::qnorm((1 + control$conf_level) / 2)
  cbind(
    n = fit$n,
    events = fit$nevent,
    loghr = estimates[, "est"],
    se = estimates[, "se"],
    ci_lower = estimates[, "est"] - q_norm * estimates[, "se"],
    ci_upper = estimates[, "est"] + q_norm * estimates[, "se"]
  )
}

#' @describeIn h_step Builds the model formula used in response STEP calculations.
#'
#' @return
#' * `h_step_rsp_formula()` returns a model formula.
#'
#' @export
h_step_rsp_formula <- function(variables,
                               control = c(control_step(), control_logistic())) {
  checkmate::assert_character(variables$covariates, null.ok = TRUE)
  assert_list_of_variables(variables[c("arm", "biomarker", "response")])
  response_definition <- sub(
    pattern = "response",
    replacement = variables$response,
    x = control$response_definition,
    fixed = TRUE
  )
  form <- paste0(response_definition, " ~ ", variables$arm)
  if (control$degree > 0) {
    form <- paste0(form, " * stats::poly(", variables$biomarker, ", degree = ", control$degree, ", raw = TRUE)")
  }
  if (!is.null(variables$covariates)) {
    form <- paste(form, "+", paste(variables$covariates, collapse = "+"))
  }
  if (!is.null(variables$strata)) {
    strata_arg <- if (length(variables$strata) > 1) {
      paste0("I(interaction(", paste0(variables$strata, collapse = ", "), "))")
    } else {
      variables$strata
    }
    form <- paste0(form, "+ strata(", strata_arg, ")")
  }
  stats::as.formula(form)
}

#' @describeIn h_step Estimates the model with `formula` built based on
#'   `variables` in `data` for a given `subset` and `control` parameters for the
#'   logistic regression.
#'
#' @param formula (`formula`)\cr the regression model formula.
#' @param subset (`logical`)\cr subset vector.
#'
#' @return
#' * `h_step_rsp_est()` returns a matrix of number of observations `n`, log odds
#'   ratio estimates `logor`, standard error `se`, and Wald confidence interval bounds
#'   `ci_lower` and `ci_upper`. One row is included for each biomarker value in `x`.
#'
#' @export
h_step_rsp_est <- function(formula,
                           data,
                           variables,
                           x,
                           subset = rep(TRUE, nrow(data)),
                           control = control_logistic()) {
  checkmate::assert_formula(formula)
  assert_df_with_variables(data, variables)
  checkmate::assert_logical(subset, min.len = 1, any.missing = FALSE)
  checkmate::assert_numeric(x, min.len = 1, any.missing = FALSE)
  checkmate::assert_list(control, names = "named")
  # Note: `subset` in `glm` needs to be an expression referring to `data` variables.
  data$.subset <- subset
  fit_warnings <- NULL
  tryCatch(
    withCallingHandlers(
      expr = {
        fit <- if (is.null(variables$strata)) {
          stats::glm(
            formula = formula,
            data = data,
            subset = .subset,
            family = stats::binomial("logit")
          )
        } else {
          # clogit needs coxph and strata imported
          survival::clogit(
            formula = formula,
            data = data,
            subset = .subset
          )
        }
      },
      warning = function(w) {
        fit_warnings <<- c(fit_warnings, w)
        invokeRestart("muffleWarning")
      }
    ),
    finally = {
    }
  )
  if (!is.null(fit_warnings)) {
    warning(paste(
      "Fit warnings occurred, please consider using a simpler model, or",
      "larger `bandwidth`, less `num_points` in `control_step()` settings"
    ))
  }
  # Produce a matrix with one row per `x` and columns `est` and `se`.
  estimates <- t(vapply(
    X = x,
    FUN = h_step_trt_effect,
    FUN.VALUE = c(1, 2),
    data = data,
    model = fit,
    variables = variables
  ))
  q_norm <- stats::qnorm((1 + control$conf_level) / 2)
  cbind(
    n = length(fit$y),
    logor = estimates[, "est"],
    se = estimates[, "se"],
    ci_lower = estimates[, "est"] - q_norm * estimates[, "se"],
    ci_upper = estimates[, "est"] + q_norm * estimates[, "se"]
  )
}

#' Confidence intervals for a difference of binomials
#'
#' @description `r lifecycle::badge("experimental")`
#'
#' Several confidence intervals for the difference between proportions.
#'
#' @name desctools_binom
NULL

#' Recycle list of parameters
#'
#' This function recycles all supplied elements to the maximal dimension.
#'
#' @param ... (`any`)\cr elements to recycle.
#'
#' @return A `list`.
#'
#' @keywords internal
#' @noRd
h_recycle <- function(...) {
  lst <- list(...)
  maxdim <- max(lengths(lst))
  res <- lapply(lst, rep, length.out = maxdim)
  attr(res, "maxdim") <- maxdim
  return(res)
}

#' @describeIn desctools_binom Several confidence intervals for the difference between proportions.
#'
#' @return A `matrix` of 3 values:
#'   * `est`: estimate of proportion difference.
#'   * `lwr.ci`: estimate of lower end of the confidence interval.
#'   * `upr.ci`: estimate of upper end of the confidence interval.
#'
#' @keywords internal
desctools_binom <- function(x1,
                            n1,
                            x2,
                            n2,
                            conf.level = 0.95, # nolint
                            sides = c("two.sided", "left", "right"),
                            method = c(
                              "ac", "wald", "waldcc", "score", "scorecc", "mn", "mee", "blj", "ha", "hal", "jp"
                            )) {
  if (missing(sides)) {
    sides <- match.arg(sides)
  }
  if (missing(method)) {
    method <- match.arg(method)
  }
  iBinomDiffCI <- function(x1, n1, x2, n2, conf.level, sides, method) { # nolint
    if (sides != "two.sided") {
      conf.level <- 1 - 2 * (1 - conf.level) # nolint
    }
    alpha <- 1 - conf.level
    kappa <- stats::qnorm(1 - alpha / 2)
    p1_hat <- x1 / n1
    p2_hat <- x2 / n2
    est <- p1_hat - p2_hat
    switch(method,
      wald = {
        vd <- p1_hat * (1 - p1_hat) / n1 + p2_hat * (1 - p2_hat) / n2
        term2 <- kappa * sqrt(vd)
        ci_lwr <- max(-1, est - term2)
        ci_upr <- min(1, est + term2)
      },
      waldcc = {
        vd <- p1_hat * (1 - p1_hat) / n1 + p2_hat * (1 - p2_hat) / n2
        term2 <- kappa * sqrt(vd)
        term2 <- term2 + 0.5 * (1 / n1 + 1 / n2)
        ci_lwr <- max(-1, est - term2)
        ci_upr <- min(1, est + term2)
      },
      ac = {
        n1 <- n1 + 2
        n2 <- n2 + 2
        x1 <- x1 + 1
        x2 <- x2 + 1
        p1_hat <- x1 / n1
        p2_hat <- x2 / n2
        est1 <- p1_hat - p2_hat
        vd <- p1_hat * (1 - p1_hat) / n1 + p2_hat * (1 - p2_hat) / n2
        term2 <- kappa * sqrt(vd)
        ci_lwr <- max(-1, est1 - term2)
        ci_upr <- min(1, est1 + term2)
      },
      exact = {
        ci_lwr <- NA
        ci_upr <- NA
      },
      score = {
        w1 <- desctools_binomci(
          x = x1, n = n1, conf.level = conf.level,
          method = "wilson"
        )
        w2 <- desctools_binomci(
          x = x2, n = n2, conf.level = conf.level,
          method = "wilson"
        )
        l1 <- w1[2]
        u1 <- w1[3]
        l2 <- w2[2]
        u2 <- w2[3]
        ci_lwr <- est - kappa * sqrt(l1 * (1 - l1) / n1 + u2 * (1 - u2) / n2)
        ci_upr <- est + kappa * sqrt(u1 * (1 - u1) / n1 + l2 * (1 - l2) / n2)
      },
      scorecc = {
        w1 <- desctools_binomci(
          x = x1, n = n1, conf.level = conf.level,
          method = "wilsoncc"
        )
        w2 <- desctools_binomci(
          x = x2, n = n2, conf.level = conf.level,
          method = "wilsoncc"
        )
        l1 <- w1[2]
        u1 <- w1[3]
        l2 <- w2[2]
        u2 <- w2[3]
        ci_lwr <- max(-1, est - sqrt((p1_hat - l1)^2 + (u2 - p2_hat)^2))
        ci_upr <- min(1, est + sqrt((u1 - p1_hat)^2 + (p2_hat - l2)^2))
      },
      mee = {
        .score <- function(p1, n1, p2, n2, dif) {
          if (dif > 1) dif <- 1
          if (dif < -1) dif <- -1
          diff <- p1 - p2 - dif
          if (abs(diff) == 0) {
            res <- 0
          } else {
            t <- n2 / n1
            a <- 1 + t
            b <- -(1 + t + p1 + t * p2 + dif * (t + 2))
            c <- dif * dif + dif * (2 * p1 + t + 1) + p1 + t * p2
            d <- -p1 * dif * (1 + dif)
            v <- (b / a / 3)^3 - b * c / (6 * a * a) + d / a / 2
            if (abs(v) < .Machine$double.eps) v <- 0
            s <- sqrt((b / a / 3)^2 - c / a / 3)
            u <- ifelse(v > 0, 1, -1) * s
            w <- (3.141592654 + acos(v / u^3)) / 3
            p1d <- 2 * u * cos(w) - b / a / 3
            p2d <- p1d - dif
            n <- n1 + n2
            res <- (p1d * (1 - p1d) / n1 + p2d * (1 - p2d) / n2)
          }
          return(sqrt(res))
        }
        pval <- function(delta) {
          z <- (est - delta) / .score(p1_hat, n1, p2_hat, n2, delta)
          2 * min(stats::pnorm(z), 1 - stats::pnorm(z))
        }
        ci_lwr <- max(-1, stats::uniroot(function(delta) {
          pval(delta) - alpha
        }, interval = c(-1 + 1e-06, est - 1e-06))$root)
        ci_upr <- min(1, stats::uniroot(function(delta) {
          pval(delta) - alpha
        }, interval = c(est + 1e-06, 1 - 1e-06))$root)
      },
      blj = {
        p1_dash <- (x1 + 0.5) / (n1 + 1)
        p2_dash <- (x2 + 0.5) / (n2 + 1)
        vd <- p1_dash * (1 - p1_dash) / n1 + p2_dash * (1 - p2_dash) / n2
        term2 <- kappa * sqrt(vd)
        est_dash <- p1_dash - p2_dash
        ci_lwr <- max(-1, est_dash - term2)
        ci_upr <- min(1, est_dash + term2)
      },
      ha = {
        term2 <- 1 /
          (2 * min(n1, n2)) + kappa * sqrt(p1_hat * (1 - p1_hat) / (n1 - 1) + p2_hat * (1 - p2_hat) / (n2 - 1))
        ci_lwr <- max(-1, est - term2)
        ci_upr <- min(1, est + term2)
      },
      mn = {
        .conf <- function(x1, n1, x2, n2, z, lower = FALSE) {
          p1 <- x1 / n1
          p2 <- x2 / n2
          p_hat <- p1 - p2
          dp <- 1 + ifelse(lower, 1, -1) * p_hat
          i <- 1
          while (i <= 50) {
            dp <- 0.5 * dp
            y <- p_hat + ifelse(lower, -1, 1) * dp
            score <- .score(p1, n1, p2, n2, y)
            if (score < z) {
              p_hat <- y
            }
            if ((dp < 1e-07) || (abs(z - score) < 1e-06)) {
              (break)()
            } else {
              i <- i + 1
            }
          }
          return(y)
        }
        .score <- function(p1, n1, p2, n2, dif) {
          diff <- p1 - p2 - dif
          if (abs(diff) == 0) {
            res <- 0
          } else {
            t <- n2 / n1
            a <- 1 + t
            b <- -(1 + t + p1 + t * p2 + dif * (t + 2))
            c <- dif * dif + dif * (2 * p1 + t + 1) + p1 + t * p2
            d <- -p1 * dif * (1 + dif)
            v <- (b / a / 3)^3 - b * c / (6 * a * a) + d / a / 2
            s <- sqrt((b / a / 3)^2 - c / a / 3)
            u <- ifelse(v > 0, 1, -1) * s
            w <- (3.141592654 + acos(v / u^3)) / 3
            p1d <- 2 * u * cos(w) - b / a / 3
            p2d <- p1d - dif
            n <- n1 + n2
            var <- (p1d * (1 - p1d) / n1 + p2d * (1 - p2d) / n2) * n / (n - 1)
            res <- diff^2 / var
          }
          return(res)
        }
        z <- stats::qchisq(conf.level, 1)
        ci_lwr <- max(-1, .conf(x1, n1, x2, n2, z, TRUE))
        ci_upr <- min(1, .conf(x1, n1, x2, n2, z, FALSE))
      },
      beal = {
        a <- p1_hat + p2_hat
        b <- p1_hat - p2_hat
        u <- ((1 / n1) + (1 / n2)) / 4
        v <- ((1 / n1) - (1 / n2)) / 4
        V <- u * ((2 - a) * a - b^2) + 2 * v * (1 - a) * b # nolint
        z <- stats::qchisq(p = 1 - alpha / 2, df = 1)
        A <- sqrt(z * (V + z * u^2 * (2 - a) * a + z * v^2 * (1 - a)^2)) # nolint
        B <- (b + z * v * (1 - a)) / (1 + z * u) # nolint
        ci_lwr <- max(-1, B - A / (1 + z * u))
        ci_upr <- min(1, B + A / (1 + z * u))
      },
      hal = {
        psi <- (p1_hat + p2_hat) / 2
        u <- (1 / n1 + 1 / n2) / 4
        v <- (1 / n1 - 1 / n2) / 4
        z <- kappa
        theta <- ((p1_hat - p2_hat) + z^2 * v * (1 - 2 * psi)) / (1 + z^2 * u)
        w <- z / (1 + z^2 * u) * sqrt(u * (4 * psi * (1 - psi) - (p1_hat - p2_hat)^2) + 2 * v * (1 - 2 * psi) *
          (p1_hat - p2_hat) + 4 * z^2 * u^2 * (1 - psi) * psi + z^2 * v^2 * (1 - 2 * psi)^2) # nolint
        c(theta + w, theta - w)
        ci_lwr <- max(-1, theta - w)
        ci_upr <- min(1, theta + w)
      },
      jp = {
        psi <- 0.5 * ((x1 + 0.5) / (n1 + 1) + (x2 + 0.5) / (n2 + 1))
        u <- (1 / n1 + 1 / n2) / 4
        v <- (1 / n1 - 1 / n2) / 4
        z <- kappa
        theta <- ((p1_hat - p2_hat) + z^2 * v * (1 - 2 * psi)) / (1 + z^2 * u)
        w <- z / (1 + z^2 * u) * sqrt(u * (4 * psi * (1 - psi) - (p1_hat - p2_hat)^2) + 2 * v * (1 - 2 * psi) *
          (p1_hat - p2_hat) + 4 * z^2 * u^2 * (1 - psi) * psi + z^2 * v^2 * (1 - 2 * psi)^2) # nolint
        c(theta + w, theta - w)
        ci_lwr <- max(-1, theta - w)
        ci_upr <- min(1, theta + w)
      },
    )
    ci <- c(
      est = est, lwr.ci = min(ci_lwr, ci_upr),
      upr.ci = max(ci_lwr, ci_upr)
    )
    if (sides == "left") {
      ci[3] <- 1
    } else if (sides == "right") {
      ci[2] <- -1
    }
    return(ci)
  }
  method <- match.arg(arg = method, several.ok = TRUE)
  sides <- match.arg(arg = sides, several.ok = TRUE)
  lst <- h_recycle(
    x1 = x1, n1 = n1, x2 = x2, n2 = n2, conf.level = conf.level,
    sides = sides, method = method
  )
  res <- t(sapply(1:attr(lst, "maxdim"), function(i) {
    iBinomDiffCI(
      x1 = lst$x1[i],
      n1 = lst$n1[i], x2 = lst$x2[i], n2 = lst$n2[i], conf.level = lst$conf.level[i],
      sides = lst$sides[i], method = lst$method[i]
    )
  }))
  lgn <- h_recycle(x1 = if (is.null(names(x1))) {
    paste("x1", seq_along(x1), sep = ".")
  } else {
    names(x1)
  }, n1 = if (is.null(names(n1))) {
    paste("n1", seq_along(n1), sep = ".")
  } else {
    names(n1)
  }, x2 = if (is.null(names(x2))) {
    paste("x2", seq_along(x2), sep = ".")
  } else {
    names(x2)
  }, n2 = if (is.null(names(n2))) {
    paste("n2", seq_along(n2), sep = ".")
  } else {
    names(n2)
  }, conf.level = conf.level, sides = sides, method = method)
  xn <- apply(as.data.frame(lgn[sapply(lgn, function(x) {
    length(unique(x)) !=
      1
  })]), 1, paste, collapse = ":")
  rownames(res) <- xn
  return(res)
}

#' @describeIn desctools_binom Compute confidence intervals for binomial proportions.
#'
#' @param x (`integer(1)`)\cr number of successes.
#' @param n (`integer(1)`)\cr number of trials.
#' @param conf.level (`proportion`)\cr confidence level, defaults to 0.95.
#' @param sides (`string`)\cr side of the confidence interval to compute. Must be one of `"two-sided"` (default),
#'   `"left"`, or `"right"`.
#' @param method (`string`)\cr method to use. Can be one out of: `"wald"`, `"wilson"`, `"wilsoncc"`,
#' `"agresti-coull"`, `"jeffreys"`, `"modified wilson"`, `"modified jeffreys"`, `"clopper-pearson"`, `"arcsine"`,
#' `"logit"`, `"witting"`, `"pratt"`, `"midp"`, `"lik"`, and `"blaker"`.
#'
#' @return A `matrix` with 3 columns containing:
#'   * `est`: estimate of proportion difference.
#'   * `lwr.ci`: lower end of the confidence interval.
#'   * `upr.ci`: upper end of the confidence interval.
#'
#' @keywords internal
desctools_binomci <- function(x,
                              n,
                              conf.level = 0.95, # nolint
                              sides = c("two.sided", "left", "right"),
                              method = c(
                                "wilson", "wald", "waldcc", "agresti-coull",
                                "jeffreys", "modified wilson", "wilsoncc", "modified jeffreys",
                                "clopper-pearson", "arcsine", "logit", "witting", "pratt",
                                "midp", "lik", "blaker"
                              ),
                              rand = 123,
                              tol = 1e-05) {
  if (missing(method)) {
    method <- "wilson"
  }
  if (missing(sides)) {
    sides <- "two.sided"
  }
  iBinomCI <- function(x, n, conf.level = 0.95, sides = c("two.sided", "left", "right"), # nolint
                       method = c(
                         "wilson", "wilsoncc", "wald",
                         "waldcc", "agresti-coull", "jeffreys", "modified wilson",
                         "modified jeffreys", "clopper-pearson", "arcsine", "logit",
                         "witting", "pratt", "midp", "lik", "blaker"
                       ),
                       rand = 123,
                       tol = 1e-05) {
    if (length(x) != 1) {
      stop("'x' has to be of length 1 (number of successes)")
    }
    if (length(n) != 1) {
      stop("'n' has to be of length 1 (number of trials)")
    }
    if (length(conf.level) != 1) {
      stop("'conf.level' has to be of length 1 (confidence level)")
    }
    if (conf.level < 0.5 || conf.level > 1) {
      stop("'conf.level' has to be in [0.5, 1]")
    }
    sides <- match.arg(sides, choices = c(
      "two.sided", "left",
      "right"
    ), several.ok = FALSE)
    if (sides != "two.sided") {
      conf.level <- 1 - 2 * (1 - conf.level) # nolint
    }
    alpha <- 1 - conf.level
    kappa <- stats::qnorm(1 - alpha / 2)
    p_hat <- x / n
    q_hat <- 1 - p_hat
    est <- p_hat
    switch(match.arg(arg = method, choices = c(
      "wilson",
      "wald", "waldcc", "wilsoncc", "agresti-coull", "jeffreys",
      "modified wilson", "modified jeffreys", "clopper-pearson",
      "arcsine", "logit", "witting", "pratt", "midp", "lik",
      "blaker"
    )),
    wald = {
      term2 <- kappa * sqrt(p_hat * q_hat) / sqrt(n)
      ci_lwr <- max(0, p_hat - term2)
      ci_upr <- min(1, p_hat + term2)
    },
    waldcc = {
      term2 <- kappa * sqrt(p_hat * q_hat) / sqrt(n)
      term2 <- term2 + 1 / (2 * n)
      ci_lwr <- max(0, p_hat - term2)
      ci_upr <- min(1, p_hat + term2)
    },
    wilson = {
      term1 <- (x + kappa^2 / 2) / (n + kappa^2)
      term2 <- kappa * sqrt(n) / (n + kappa^2) * sqrt(p_hat * q_hat + kappa^2 / (4 * n))
      ci_lwr <- max(0, term1 - term2)
      ci_upr <- min(1, term1 + term2)
    },
    wilsoncc = {
      lci <- (
        2 * x + kappa^2 - 1 - kappa * sqrt(kappa^2 - 2 - 1 / n + 4 * p_hat * (n * q_hat + 1))
      ) / (2 * (n + kappa^2))
      uci <- (
        2 * x + kappa^2 + 1 + kappa * sqrt(kappa^2 + 2 - 1 / n + 4 * p_hat * (n * q_hat - 1))
      ) / (2 * (n + kappa^2))
      ci_lwr <- max(0, ifelse(p_hat == 0, 0, lci))
      ci_upr <- min(1, ifelse(p_hat == 1, 1, uci))
    },
    `agresti-coull` = {
      x_tilde <- x + kappa^2 / 2
      n_tilde <- n + kappa^2
      p_tilde <- x_tilde / n_tilde
      q_tilde <- 1 - p_tilde
      est <- p_tilde
      term2 <- kappa * sqrt(p_tilde * q_tilde) / sqrt(n_tilde)
      ci_lwr <- max(0, p_tilde - term2)
      ci_upr <- min(1, p_tilde + term2)
    },
    jeffreys = {
      if (x == 0) {
        ci_lwr <- 0
      } else {
        ci_lwr <- stats::qbeta(
          alpha / 2,
          x + 0.5, n - x + 0.5
        )
      }
      if (x == n) {
        ci_upr <- 1
      } else {
        ci_upr <- stats::qbeta(1 - alpha / 2, x + 0.5, n - x + 0.5)
      }
    },
    `modified wilson` = {
      term1 <- (x + kappa^2 / 2) / (n + kappa^2)
      term2 <- kappa * sqrt(n) / (n + kappa^2) * sqrt(p_hat * q_hat + kappa^2 / (4 * n))
      if ((n <= 50 & x %in% c(1, 2)) | (n >= 51 & x %in% c(1:3))) {
        ci_lwr <- 0.5 * stats::qchisq(alpha, 2 * x) / n
      } else {
        ci_lwr <- max(0, term1 - term2)
      }
      if ((n <= 50 & x %in% c(n - 1, n - 2)) | (n >= 51 & x %in% c(n - (1:3)))) {
        ci_upr <- 1 - 0.5 * stats::qchisq(
          alpha,
          2 * (n - x)
        ) / n
      } else {
        ci_upr <- min(1, term1 + term2)
      }
    },
    `modified jeffreys` = {
      if (x == n) {
        ci_lwr <- (alpha / 2)^(1 / n)
      } else {
        if (x <= 1) {
          ci_lwr <- 0
        } else {
          ci_lwr <- stats::qbeta(
            alpha / 2,
            x + 0.5, n - x + 0.5
          )
        }
      }
      if (x == 0) {
        ci_upr <- 1 - (alpha / 2)^(1 / n)
      } else {
        if (x >= n - 1) {
          ci_upr <- 1
        } else {
          ci_upr <- stats::qbeta(1 - alpha / 2, x + 0.5, n - x + 0.5)
        }
      }
    },
    `clopper-pearson` = {
      ci_lwr <- stats::qbeta(alpha / 2, x, n - x + 1)
      ci_upr <- stats::qbeta(1 - alpha / 2, x + 1, n - x)
    },
    arcsine = {
      p_tilde <- (x + 0.375) / (n + 0.75)
      est <- p_tilde
      ci_lwr <- sin(asin(sqrt(p_tilde)) - 0.5 * kappa / sqrt(n))^2
      ci_upr <- sin(asin(sqrt(p_tilde)) + 0.5 * kappa / sqrt(n))^2
    },
    logit = {
      lambda_hat <- log(x / (n - x))
      V_hat <- n / (x * (n - x)) # nolint
      lambda_lower <- lambda_hat - kappa * sqrt(V_hat)
      lambda_upper <- lambda_hat + kappa * sqrt(V_hat)
      ci_lwr <- exp(lambda_lower) / (1 + exp(lambda_lower))
      ci_upr <- exp(lambda_upper) / (1 + exp(lambda_upper))
    },
    witting = {
      set.seed(rand)
      x_tilde <- x + stats::runif(1, min = 0, max = 1)
      pbinom_abscont <- function(q, size, prob) {
        v <- trunc(q)
        term1 <- stats::pbinom(v - 1, size = size, prob = prob)
        term2 <- (q - v) * stats::dbinom(v, size = size, prob = prob)
        return(term1 + term2)
      }
      qbinom_abscont <- function(p, size, x) {
        fun <- function(prob, size, x, p) {
          pbinom_abscont(x, size, prob) - p
        }
        stats::uniroot(fun,
          interval = c(0, 1), size = size,
          x = x, p = p
        )$root
      }
      ci_lwr <- qbinom_abscont(1 - alpha, size = n, x = x_tilde)
      ci_upr <- qbinom_abscont(alpha, size = n, x = x_tilde)
    },
    pratt = {
      if (x == 0) {
        ci_lwr <- 0
        ci_upr <- 1 - alpha^(1 / n)
      } else if (x == 1) {
        ci_lwr <- 1 - (1 - alpha / 2)^(1 / n)
        ci_upr <- 1 - (alpha / 2)^(1 / n)
      } else if (x == (n - 1)) {
        ci_lwr <- (alpha / 2)^(1 / n)
        ci_upr <- (1 - alpha / 2)^(1 / n)
      } else if (x == n) {
        ci_lwr <- alpha^(1 / n)
        ci_upr <- 1
      } else {
        z <- stats::qnorm(1 - alpha / 2)
        A <- ((x + 1) / (n - x))^2 # nolint
        B <- 81 * (x + 1) * (n - x) - 9 * n - 8 # nolint
        C <- (0 - 3) * z * sqrt(9 * (x + 1) * (n - x) * (9 * n + 5 - z^2) + n + 1) # nolint
        D <- 81 * (x + 1)^2 - 9 * (x + 1) * (2 + z^2) + 1 # nolint
        E <- 1 + A * ((B + C) / D)^3 # nolint
        ci_upr <- 1 / E
        A <- (x / (n - x - 1))^2 # nolint
        B <- 81 * x * (n - x - 1) - 9 * n - 8 # nolint
        C <- 3 * z * sqrt(9 * x * (n - x - 1) * (9 * n + 5 - z^2) + n + 1) # nolint
        D <- 81 * x^2 - 9 * x * (2 + z^2) + 1 # nolint
        E <- 1 + A * ((B + C) / D)^3 # nolint
        ci_lwr <- 1 / E
      }
    },
    midp = {
      f_low <- function(pi, x, n) {
        1 / 2 * stats::dbinom(x, size = n, prob = pi) + stats::pbinom(x,
          size = n, prob = pi, lower.tail = FALSE
        ) -
          (1 - conf.level) / 2
      }
      f_up <- function(pi, x, n) {
        1 / 2 * stats::dbinom(x, size = n, prob = pi) + stats::pbinom(x - 1, size = n, prob = pi) - (1 - conf.level) / 2
      }
      ci_lwr <- 0
      ci_upr <- 1
      if (x != 0) {
        ci_lwr <- stats::uniroot(f_low,
          interval = c(0, p_hat),
          x = x, n = n
        )$root
      }
      if (x != n) {
        ci_upr <- stats::uniroot(f_up, interval = c(
          p_hat,
          1
        ), x = x, n = n)$root
      }
    },
    lik = {
      ci_lwr <- 0
      ci_upr <- 1
      z <- stats::qnorm(1 - alpha * 0.5)
      tol <- .Machine$double.eps^0.5
      BinDev <- function(y, x, mu, wt, bound = 0, tol = .Machine$double.eps^0.5, # nolint
                         ...) {
        ll_y <- ifelse(y %in% c(0, 1), 0, stats::dbinom(x, wt,
          y,
          log = TRUE
        ))
        ll_mu <- ifelse(mu %in% c(0, 1), 0, stats::dbinom(x,
          wt, mu,
          log = TRUE
        ))
        res <- ifelse(abs(y - mu) < tol, 0, sign(y - mu) * sqrt(-2 * (ll_y - ll_mu)))
        return(res - bound)
      }
      if (x != 0 && tol < p_hat) {
        ci_lwr <- if (BinDev(
          tol, x, p_hat, n, -z,
          tol
        ) <= 0) {
          stats::uniroot(
            f = BinDev, interval = c(tol, if (p_hat < tol || p_hat == 1) {
              1 - tol
            } else {
              p_hat
            }), bound = -z,
            x = x, mu = p_hat, wt = n
          )$root
        }
      }
      if (x != n && p_hat < (1 - tol)) {
        ci_upr <- if (
          BinDev(y = 1 - tol, x = x, mu = ifelse(p_hat > 1 - tol, tol, p_hat), wt = n, bound = z, tol = tol) < 0) { # nolint
          ci_lwr <- if (BinDev(
            tol, x, if (p_hat < tol || p_hat == 1) {
              1 - tol
            } else {
              p_hat
            }, n,
            -z, tol
          ) <= 0) {
            stats::uniroot(
              f = BinDev, interval = c(tol, p_hat),
              bound = -z, x = x, mu = p_hat, wt = n
            )$root
          }
        } else {
          stats::uniroot(
            f = BinDev, interval = c(if (p_hat > 1 - tol) {
              tol
            } else {
              p_hat
            }, 1 - tol), bound = z,
            x = x, mu = p_hat, wt = n
          )$root
        }
      }
    },
    blaker = {
      acceptbin <- function(x, n, p) {
        p1 <- 1 - stats::pbinom(x - 1, n, p)
        p2 <- stats::pbinom(x, n, p)
        a1 <- p1 + stats::pbinom(stats::qbinom(p1, n, p) - 1, n, p)
        a2 <- p2 + 1 - stats::pbinom(
          stats::qbinom(1 - p2, n, p), n,
          p
        )
        return(min(a1, a2))
      }
      ci_lwr <- 0
      ci_upr <- 1
      if (x != 0) {
        ci_lwr <- stats::qbeta((1 - conf.level) / 2, x, n - x + 1)
        while (acceptbin(x, n, ci_lwr + tol) < (1 - conf.level)) {
          ci_lwr <- ci_lwr + tol
        }
      }
      if (x != n) {
        ci_upr <- stats::qbeta(1 - (1 - conf.level) / 2, x + 1, n - x)
        while (acceptbin(x, n, ci_upr - tol) < (1 - conf.level)) {
          ci_upr <- ci_upr - tol
        }
      }
    }
    )
    ci <- c(est = est, lwr.ci = max(0, ci_lwr), upr.ci = min(
      1,
      ci_upr
    ))
    if (sides == "left") {
      ci[3] <- 1
    } else if (sides == "right") {
      ci[2] <- 0
    }
    return(ci)
  }
  lst <- list(
    x = x, n = n, conf.level = conf.level, sides = sides,
    method = method, rand = rand
  )
  maxdim <- max(unlist(lapply(lst, length)))
  lgp <- lapply(lst, rep, length.out = maxdim)
  lgn <- h_recycle(x = if (is.null(names(x))) {
    paste("x", seq_along(x), sep = ".")
  } else {
    names(x)
  }, n = if (is.null(names(n))) {
    paste("n", seq_along(n), sep = ".")
  } else {
    names(n)
  }, conf.level = conf.level, sides = sides, method = method)
  xn <- apply(as.data.frame(lgn[sapply(lgn, function(x) {
    length(unique(x)) !=
      1
  })]), 1, paste, collapse = ":")
  res <- t(sapply(1:maxdim, function(i) {
    iBinomCI(
      x = lgp$x[i],
      n = lgp$n[i], conf.level = lgp$conf.level[i], sides = lgp$sides[i],
      method = lgp$method[i], rand = lgp$rand[i]
    )
  }))
  colnames(res)[1] <- c("est")
  rownames(res) <- xn
  return(res)
}

#' Tabulate survival duration by subgroup
#'
#' @description `r lifecycle::badge("stable")`
#'
#' The [tabulate_survival_subgroups()] function creates a layout element to tabulate survival duration by subgroup,
#' returning statistics including median survival time and hazard ratio for each population subgroup. The table is
#' created from `df`, a list of data frames returned by [extract_survival_subgroups()], with the statistics to include
#' specified via the `vars` parameter.
#'
#' A forest plot can be created from the resulting table using the [g_forest()] function.
#'
#' @inheritParams argument_convention
#' @inheritParams survival_coxph_pairwise
#' @param df (`list`)\cr list of data frames containing all analysis variables. List should be
#'   created using [extract_survival_subgroups()].
#' @param vars (`character`)\cr the names of statistics to be reported among:
#'   * `n_tot_events`: Total number of events per group.
#'   * `n_events`: Number of events per group.
#'   * `n_tot`: Total number of observations per group.
#'   * `n`: Number of observations per group.
#'   * `median`: Median survival time.
#'   * `hr`: Hazard ratio.
#'   * `ci`: Confidence interval of hazard ratio.
#'   * `pval`: p-value of the effect.
#'   Note, one of the statistics `n_tot` and `n_tot_events`, as well as both `hr` and `ci`
#'   are required.
#' @param time_unit (`string`)\cr label with unit of median survival time. Default `NULL` skips displaying unit.
#'
#' @details These functions create a layout starting from a data frame which contains
#'   the required statistics. Tables typically used as part of forest plot.
#'
#' @seealso [extract_survival_subgroups()]
#'
#' @examples
#' library(dplyr)
#'
#' adtte <- tern_ex_adtte
#'
#' # Save variable labels before data processing steps.
#' adtte_labels <- formatters::var_labels(adtte)
#'
#' adtte_f <- adtte %>%
#'   filter(
#'     PARAMCD == "OS",
#'     ARM %in% c("B: Placebo", "A: Drug X"),
#'     SEX %in% c("M", "F")
#'   ) %>%
#'   mutate(
#'     # Reorder levels of ARM to display reference arm before treatment arm.
#'     ARM = droplevels(forcats::fct_relevel(ARM, "B: Placebo")),
#'     SEX = droplevels(SEX),
#'     AVALU = as.character(AVALU),
#'     is_event = CNSR == 0
#'   )
#' labels <- c(
#'   "ARM" = adtte_labels[["ARM"]],
#'   "SEX" = adtte_labels[["SEX"]],
#'   "AVALU" = adtte_labels[["AVALU"]],
#'   "is_event" = "Event Flag"
#' )
#' formatters::var_labels(adtte_f)[names(labels)] <- labels
#'
#' df <- extract_survival_subgroups(
#'   variables = list(
#'     tte = "AVAL",
#'     is_event = "is_event",
#'     arm = "ARM", subgroups = c("SEX", "BMRKR2")
#'   ),
#'   label_all = "Total Patients",
#'   data = adtte_f
#' )
#' df
#'
#' df_grouped <- extract_survival_subgroups(
#'   variables = list(
#'     tte = "AVAL",
#'     is_event = "is_event",
#'     arm = "ARM", subgroups = c("SEX", "BMRKR2")
#'   ),
#'   data = adtte_f,
#'   groups_lists = list(
#'     BMRKR2 = list(
#'       "low" = "LOW",
#'       "low/medium" = c("LOW", "MEDIUM"),
#'       "low/medium/high" = c("LOW", "MEDIUM", "HIGH")
#'     )
#'   )
#' )
#' df_grouped
#'
#' @name survival_duration_subgroups
#' @order 1
NULL

#' Prepare survival data for population subgroups in data frames
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Prepares estimates of median survival times and treatment hazard ratios for population subgroups in
#' data frames. Simple wrapper for [h_survtime_subgroups_df()] and [h_coxph_subgroups_df()]. Result is a `list`
#' of two `data.frame`s: `survtime` and `hr`. `variables` corresponds to the names of variables found in `data`,
#' passed as a named `list` and requires elements `tte`, `is_event`, `arm` and optionally `subgroups` and `strata`.
#' `groups_lists` optionally specifies groupings for `subgroups` variables.
#'
#' @inheritParams argument_convention
#' @inheritParams survival_duration_subgroups
#' @inheritParams survival_coxph_pairwise
#'
#' @return A named `list` of two elements:
#'   * `survtime`: A `data.frame` containing columns `arm`, `n`, `n_events`, `median`, `subgroup`, `var`,
#'     `var_label`, and `row_type`.
#'   * `hr`: A `data.frame` containing columns `arm`, `n_tot`, `n_tot_events`, `hr`, `lcl`, `ucl`, `conf_level`,
#'     `pval`, `pval_label`, `subgroup`, `var`, `var_label`, and `row_type`.
#'
#' @seealso [survival_duration_subgroups]
#'
#' @export
extract_survival_subgroups <- function(variables,
                                       data,
                                       groups_lists = list(),
                                       control = control_coxph(),
                                       label_all = "All Patients") {
  if ("strat" %in% names(variables)) {
    warning(
      "Warning: the `strat` element name of the `variables` list argument to `extract_survival_subgroups() ",
      "was deprecated in tern 0.9.4.\n  ",
      "Please use the name `strata` instead of `strat` in the `variables` argument."
    )
    variables[["strata"]] <- variables[["strat"]]
  }

  df_survtime <- h_survtime_subgroups_df(
    variables,
    data,
    groups_lists = groups_lists,
    label_all = label_all
  )
  df_hr <- h_coxph_subgroups_df(
    variables,
    data,
    groups_lists = groups_lists,
    control = control,
    label_all = label_all
  )

  list(survtime = df_survtime, hr = df_hr)
}

#' @describeIn survival_duration_subgroups  Formatted analysis function which is used as
#'   `afun` in `tabulate_survival_subgroups()`.
#'
#' @return
#' * `a_survival_subgroups()` returns the corresponding list with formatted [rtables::CellValue()].
#'
#' @keywords internal
a_survival_subgroups <- function(.formats = list( # nolint start
                                   n = "xx",
                                   n_events = "xx",
                                   n_tot_events = "xx",
                                   median = "xx.x",
                                   n_tot = "xx",
                                   hr = list(format_extreme_values(2L)),
                                   ci = list(format_extreme_values_ci(2L)),
                                   pval = "x.xxxx | (<0.0001)"
                                 ),
                                 na_str = default_na_str()) { # nolint end
  checkmate::assert_list(.formats)
  checkmate::assert_subset(
    names(.formats),
    c("n", "n_events", "median", "n_tot", "n_tot_events", "hr", "ci", "pval", "riskdiff")
  )

  afun_lst <- Map(
    function(stat, fmt, na_str) {
      function(df, labelstr = "", ...) {
        in_rows(
          .list = as.list(df[[stat]]),
          .labels = as.character(df$subgroup),
          .formats = fmt,
          .format_na_strs = na_str
        )
      }
    },
    stat = names(.formats),
    fmt = .formats,
    na_str = na_str
  )

  afun_lst
}

#' @describeIn survival_duration_subgroups Table-creating function which creates a table
#'   summarizing survival by subgroup. This function is a wrapper for [rtables::analyze_colvars()]
#'   and [rtables::summarize_row_groups()].
#'
#' @param label_all `r lifecycle::badge("deprecated")`\cr please assign the `label_all` parameter within the
#'   [extract_survival_subgroups()] function when creating `df`.
#' @param riskdiff (`list`)\cr if a risk (proportion) difference column should be added, a list of settings to apply
#'   within the column. See [control_riskdiff()] for details. If `NULL`, no risk difference column will be added. If
#'   `riskdiff$arm_x` and `riskdiff$arm_y` are `NULL`, the first level of `df$survtime$arm` will be used as `arm_x`
#'   and the second level as `arm_y`.
#'
#' @return An `rtables` table summarizing survival by subgroup.
#'
#' @examples
#' ## Table with default columns.
#' basic_table() %>%
#'   tabulate_survival_subgroups(df, time_unit = adtte_f$AVALU[1])
#'
#' ## Table with a manually chosen set of columns: adding "pval".
#' basic_table() %>%
#'   tabulate_survival_subgroups(
#'     df = df,
#'     vars = c("n_tot_events", "n_events", "median", "hr", "ci", "pval"),
#'     time_unit = adtte_f$AVALU[1]
#'   )
#'
#' @export
#' @order 2
tabulate_survival_subgroups <- function(lyt,
                                        df,
                                        vars = c("n_tot_events", "n_events", "median", "hr", "ci"),
                                        groups_lists = list(),
                                        label_all = lifecycle::deprecated(),
                                        time_unit = NULL,
                                        riskdiff = NULL,
                                        na_str = default_na_str(),
                                        .formats = c(
                                          n = "xx", n_events = "xx", n_tot_events = "xx", median = "xx.x", n_tot = "xx",
                                          hr = list(format_extreme_values(2L)), ci = list(format_extreme_values_ci(2L)),
                                          pval = "x.xxxx | (<0.0001)"
                                        )) {
  checkmate::assert_list(riskdiff, null.ok = TRUE)
  checkmate::assert_true(any(c("n_tot", "n_tot_events") %in% vars))
  checkmate::assert_true(all(c("hr", "ci") %in% vars))

  if (lifecycle::is_present(label_all)) {
    lifecycle::deprecate_warn(
      "0.9.5", "tabulate_survival_subgroups(label_all)",
      details =
        "Please assign the `label_all` parameter within the `extract_survival_subgroups()` function when creating `df`."
    )
  }

  # Create "ci" column from "lcl" and "ucl"
  df$hr$ci <- combine_vectors(df$hr$lcl, df$hr$ucl)

  # Fill in missing formats with defaults
  default_fmts <- eval(formals(tabulate_survival_subgroups)$.formats)
  .formats <- c(.formats, default_fmts[vars[!vars %in% names(.formats)]])

  # Extract additional parameters from df
  conf_level <- df$hr$conf_level[1]
  method <- df$hr$pval_label[1]
  colvars <- d_survival_subgroups_colvars(vars, conf_level = conf_level, method = method, time_unit = time_unit)
  survtime_vars <- intersect(colvars$vars, c("n", "n_events", "median"))
  hr_vars <- intersect(names(colvars$labels), c("n_tot", "n_tot_events", "hr", "ci", "pval"))
  colvars_survtime <- list(vars = survtime_vars, labels = colvars$labels[survtime_vars])
  colvars_hr <- list(vars = hr_vars, labels = colvars$labels[hr_vars])

  extra_args <- list(groups_lists = groups_lists, conf_level = conf_level, method = method)

  # Get analysis function for each statistic
  afun_lst <- a_survival_subgroups(.formats = c(.formats, riskdiff = riskdiff$format), na_str = na_str)

  # Add risk difference column
  if (!is.null(riskdiff)) {
    if (is.null(riskdiff$arm_x)) riskdiff$arm_x <- levels(df$survtime$arm)[1]
    if (is.null(riskdiff$arm_y)) riskdiff$arm_y <- levels(df$survtime$arm)[2]
    colvars_hr$vars <- c(colvars_hr$vars, "riskdiff")
    colvars_hr$labels <- c(colvars_hr$labels, riskdiff = riskdiff$col_label)
    arm_cols <- paste(rep(c("n_events", "n_events", "n", "n")), c(riskdiff$arm_x, riskdiff$arm_y), sep = "_")

    df_prop_diff <- df$survtime %>%
      dplyr::select(-"median") %>%
      tidyr::pivot_wider(
        id_cols = c("subgroup", "var", "var_label", "row_type"),
        names_from = "arm",
        values_from = c("n", "n_events")
      ) %>%
      dplyr::rowwise() %>%
      dplyr::mutate(
        riskdiff = stat_propdiff_ci(
          x = as.list(.data[[arm_cols[1]]]),
          y = as.list(.data[[arm_cols[2]]]),
          N_x = .data[[arm_cols[3]]],
          N_y = .data[[arm_cols[4]]]
        )
      ) %>%
      dplyr::select(-dplyr::all_of(arm_cols))

    df$hr <- df$hr %>%
      dplyr::left_join(
        df_prop_diff,
        by = c("subgroup", "var", "var_label", "row_type")
      )
  }

  # Add columns from table_survtime (optional)
  if (length(colvars_survtime$vars) > 0) {
    lyt_survtime <- split_cols_by(lyt = lyt, var = "arm")
    lyt_survtime <- split_rows_by(
      lyt = lyt_survtime,
      var = "row_type",
      split_fun = keep_split_levels("content"),
      nested = FALSE
    )

    # Add "All Patients" row
    lyt_survtime <- summarize_row_groups(
      lyt = lyt_survtime,
      var = "var_label",
      cfun = afun_lst[names(colvars_survtime$labels)],
      na_str = na_str,
      extra_args = extra_args
    )
    lyt_survtime <- split_cols_by_multivar(
      lyt = lyt_survtime,
      vars = colvars_survtime$vars,
      varlabels = colvars_survtime$labels
    )

    # Add analysis rows
    if ("analysis" %in% df$survtime$row_type) {
      lyt_survtime <- split_rows_by(
        lyt = lyt_survtime,
        var = "row_type",
        split_fun = keep_split_levels("analysis"),
        nested = FALSE,
        child_labels = "hidden"
      )
      lyt_survtime <- split_rows_by(lyt = lyt_survtime, var = "var_label", nested = TRUE)
      lyt_survtime <- analyze_colvars(
        lyt = lyt_survtime,
        afun = afun_lst[names(colvars_survtime$labels)],
        na_str = na_str,
        inclNAs = TRUE,
        extra_args = extra_args
      )
    }

    table_survtime <- build_table(lyt_survtime, df = df$survtime)
  } else {
    table_survtime <- NULL
  }

  # Add columns from table_hr ("n_tot_events" or "n_tot", "or" and "ci" required)
  lyt_hr <- split_cols_by(lyt = lyt, var = "arm")
  lyt_hr <- split_rows_by(
    lyt = lyt_hr,
    var = "row_type",
    split_fun = keep_split_levels("content"),
    nested = FALSE
  )
  lyt_hr <- summarize_row_groups(
    lyt = lyt_hr,
    var = "var_label",
    cfun = afun_lst[names(colvars_hr$labels)],
    na_str = na_str,
    extra_args = extra_args
  )
  lyt_hr <- split_cols_by_multivar(
    lyt = lyt_hr,
    vars = colvars_hr$vars,
    varlabels = colvars_hr$labels
  ) %>%
    append_topleft("Baseline Risk Factors")

  # Add analysis rows
  if ("analysis" %in% df$survtime$row_type) {
    lyt_hr <- split_rows_by(
      lyt = lyt_hr,
      var = "row_type",
      split_fun = keep_split_levels("analysis"),
      nested = FALSE,
      child_labels = "hidden"
    )
    lyt_hr <- split_rows_by(lyt = lyt_hr, var = "var_label", nested = TRUE)
    lyt_hr <- analyze_colvars(
      lyt = lyt_hr,
      afun = afun_lst[names(colvars_hr$labels)],
      na_str = na_str,
      inclNAs = TRUE,
      extra_args = extra_args
    )
  }

  table_hr <- build_table(lyt_hr, df = df$hr)

  # Join tables, add forest plot attributes
  n_tot_ids <- grep("^n_tot", colvars_hr$vars)
  if (is.null(table_survtime)) {
    result <- table_hr
    hr_id <- match("hr", colvars_hr$vars)
    ci_id <- match("ci", colvars_hr$vars)
  } else {
    result <- cbind_rtables(table_hr[, n_tot_ids], table_survtime, table_hr[, -n_tot_ids])
    hr_id <- length(n_tot_ids) + ncol(table_survtime) + match("hr", colvars_hr$vars[-n_tot_ids])
    ci_id <- length(n_tot_ids) + ncol(table_survtime) + match("ci", colvars_hr$vars[-n_tot_ids])
    n_tot_ids <- seq_along(n_tot_ids)
  }
  structure(
    result,
    forest_header = paste0(rev(levels(df$survtime$arm)), "\nBetter"),
    col_x = hr_id,
    col_ci = ci_id,
    col_symbol_size = n_tot_ids[1] # for scaling the symbol sizes in forest plots
  )
}

#' Labels for column variables in survival duration by subgroup table
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Internal function to check variables included in [tabulate_survival_subgroups()] and create column labels.
#'
#' @inheritParams tabulate_survival_subgroups
#' @inheritParams argument_convention
#' @param method (`string`)\cr p-value method for testing hazard ratio = 1.
#'
#' @return A `list` of variables and their labels to tabulate.
#'
#' @note At least one of `n_tot` and `n_tot_events` must be provided in `vars`.
#'
#' @export
d_survival_subgroups_colvars <- function(vars,
                                         conf_level,
                                         method,
                                         time_unit = NULL) {
  checkmate::assert_character(vars)
  checkmate::assert_string(time_unit, null.ok = TRUE)
  checkmate::assert_subset(c("hr", "ci"), vars)
  checkmate::assert_true(any(c("n_tot", "n_tot_events") %in% vars))
  checkmate::assert_subset(
    vars,
    c("n", "n_events", "median", "n_tot", "n_tot_events", "hr", "ci", "pval")
  )

  propcase_time_label <- if (!is.null(time_unit)) {
    paste0("Median (", time_unit, ")")
  } else {
    "Median"
  }

  varlabels <- c(
    n = "n",
    n_events = "Events",
    median = propcase_time_label,
    n_tot = "Total n",
    n_tot_events = "Total Events",
    hr = "Hazard Ratio",
    ci = paste0(100 * conf_level, "% Wald CI"),
    pval = method
  )

  colvars <- vars

  # The `lcl` variable is just a placeholder available in the analysis data,
  # it is not acutally used in the tabulation.
  # Variables used in the tabulation are lcl and ucl, see `a_survival_subgroups` for details.
  colvars[colvars == "ci"] <- "lcl"

  list(
    vars = colvars,
    labels = varlabels[vars]
  )
}

#' Helper functions for incidence rate
#'
#' @description `r lifecycle::badge("stable")`
#'
#' @param control (`list`)\cr parameters for estimation details, specified by using
#'   the helper function [control_incidence_rate()]. Possible parameter options are:
#'   * `conf_level`: (`proportion`)\cr confidence level for the estimated incidence rate.
#'   * `conf_type`: (`string`)\cr `normal` (default), `normal_log`, `exact`, or `byar`
#'     for confidence interval type.
#'   * `input_time_unit`: (`string`)\cr `day`, `week`, `month`, or `year` (default)
#'     indicating time unit for data input.
#'   * `num_pt_year`: (`numeric`)\cr time unit for desired output (in person-years).
#' @param person_years (`numeric(1)`)\cr total person-years at risk.
#' @param alpha (`numeric(1)`)\cr two-sided alpha-level for confidence interval.
#' @param n_events (`integer(1)`)\cr number of events observed.
#'
#' @return Estimated incidence rate, `rate`, and associated confidence interval, `rate_ci`.
#'
#' @seealso [incidence_rate]
#'
#' @name h_incidence_rate
NULL

#' @describeIn h_incidence_rate Helper function to estimate the incidence rate and
#'   associated confidence interval.
#'
#' @keywords internal
h_incidence_rate <- function(person_years,
                             n_events,
                             control = control_incidence_rate()) {
  alpha <- 1 - control$conf_level
  est <- switch(control$conf_type,
    normal = h_incidence_rate_normal(person_years, n_events, alpha),
    normal_log = h_incidence_rate_normal_log(person_years, n_events, alpha),
    exact = h_incidence_rate_exact(person_years, n_events, alpha),
    byar = h_incidence_rate_byar(person_years, n_events, alpha)
  )

  num_pt_year <- control$num_pt_year
  list(
    rate = est$rate * num_pt_year,
    rate_ci = est$rate_ci * num_pt_year
  )
}

#' @describeIn h_incidence_rate Helper function to estimate the incidence rate and
#'   associated confidence interval based on the normal approximation for the
#'   incidence rate. Unit is one person-year.
#'
#' @examples
#' h_incidence_rate_normal(200, 2)
#'
#' @export
h_incidence_rate_normal <- function(person_years,
                                    n_events,
                                    alpha = 0.05) {
  checkmate::assert_number(person_years)
  checkmate::assert_number(n_events)
  assert_proportion_value(alpha)

  est <- n_events / person_years
  se <- sqrt(est / person_years)
  ci <- est + c(-1, 1) * stats::qnorm(1 - alpha / 2) * se

  list(rate = est, rate_ci = ci)
}

#' @describeIn h_incidence_rate Helper function to estimate the incidence rate and
#'   associated confidence interval based on the normal approximation for the
#'   logarithm of the incidence rate. Unit is one person-year.
#'
#' @examples
#' h_incidence_rate_normal_log(200, 2)
#'
#' @export
h_incidence_rate_normal_log <- function(person_years,
                                        n_events,
                                        alpha = 0.05) {
  checkmate::assert_number(person_years)
  checkmate::assert_number(n_events)
  assert_proportion_value(alpha)

  rate_est <- n_events / person_years
  rate_se <- sqrt(rate_est / person_years)
  lrate_est <- log(rate_est)
  lrate_se <- rate_se / rate_est
  ci <- exp(lrate_est + c(-1, 1) * stats::qnorm(1 - alpha / 2) * lrate_se)

  list(rate = rate_est, rate_ci = ci)
}

#' @describeIn h_incidence_rate Helper function to estimate the incidence rate and
#'   associated exact confidence interval. Unit is one person-year.
#'
#' @examples
#' h_incidence_rate_exact(200, 2)
#'
#' @export
h_incidence_rate_exact <- function(person_years,
                                   n_events,
                                   alpha = 0.05) {
  checkmate::assert_number(person_years)
  checkmate::assert_number(n_events)
  assert_proportion_value(alpha)

  est <- n_events / person_years
  lcl <- stats::qchisq(p = (alpha) / 2, df = 2 * n_events) / (2 * person_years)
  ucl <- stats::qchisq(p = 1 - (alpha) / 2, df = 2 * n_events + 2) / (2 * person_years)

  list(rate = est, rate_ci = c(lcl, ucl))
}

#' @describeIn h_incidence_rate Helper function to estimate the incidence rate and
#'   associated Byar's confidence interval. Unit is one person-year.
#'
#' @examples
#' h_incidence_rate_byar(200, 2)
#'
#' @export
h_incidence_rate_byar <- function(person_years,
                                  n_events,
                                  alpha = 0.05) {
  checkmate::assert_number(person_years)
  checkmate::assert_number(n_events)
  assert_proportion_value(alpha)

  est <- n_events / person_years
  seg_1 <- n_events + 0.5
  seg_2 <- 1 - 1 / (9 * (n_events + 0.5))
  seg_3 <- stats::qnorm(1 - alpha / 2) * sqrt(1 / (n_events + 0.5)) / 3
  lcl <- seg_1 * ((seg_2 - seg_3)^3) / person_years
  ucl <- seg_1 * ((seg_2 + seg_3)^3) / person_years

  list(rate = est, rate_ci = c(lcl, ucl))
}

#' Individual patient plots
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Line plot(s) displaying trend in patients' parameter values over time is rendered.
#' Patients' individual baseline values can be added to the plot(s) as reference.
#'
#' @inheritParams argument_convention
#' @param xvar (`string`)\cr time point variable to be plotted on x-axis.
#' @param yvar (`string`)\cr continuous analysis variable to be plotted on y-axis.
#' @param xlab (`string`)\cr plot label for x-axis.
#' @param ylab (`string`)\cr plot label for y-axis.
#' @param id_var (`string`)\cr variable used as patient identifier.
#' @param title (`string`)\cr title for plot.
#' @param subtitle (`string`)\cr subtitle for plot.
#' @param add_baseline_hline (`flag`)\cr adds horizontal line at baseline y-value on
#'   plot when `TRUE`.
#' @param yvar_baseline (`string`)\cr variable with baseline values only.
#'   Ignored when `add_baseline_hline` is `FALSE`.
#' @param ggtheme (`theme`)\cr optional graphical theme function as provided
#'   by `ggplot2` to control outlook of plot. Use `ggplot2::theme()` to tweak the display.
#' @param plotting_choices (`string`)\cr specifies options for displaying
#'   plots. Must be one of `"all_in_one"`, `"split_by_max_obs"`, or `"separate_by_obs"`.
#' @param max_obs_per_plot (`integer(1)`)\cr number of observations to be plotted on one
#'   plot. Ignored if `plotting_choices` is not `"separate_by_obs"`.
#' @param caption (`string`)\cr optional caption below the plot.
#' @param col (`character`)\cr line colors.
#'
#' @seealso Relevant helper function [h_g_ipp()].
#'
#' @name g_ipp
#' @aliases individual_patient_plot
NULL

#' Helper function to create simple line plot over time
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Function that generates a simple line plot displaying parameter trends over time.
#'
#' @inheritParams argument_convention
#' @inheritParams g_ipp
#'
#' @return A `ggplot` line plot.
#'
#' @seealso [g_ipp()] which uses this function.
#'
#' @examples
#' library(dplyr)
#' library(nestcolor)
#'
#' # Select a small sample of data to plot.
#' adlb <- tern_ex_adlb %>%
#'   filter(PARAMCD == "ALT", !(AVISIT %in% c("SCREENING", "BASELINE"))) %>%
#'   slice(1:36)
#'
#' p <- h_g_ipp(
#'   df = adlb,
#'   xvar = "AVISIT",
#'   yvar = "AVAL",
#'   xlab = "Visit",
#'   id_var = "USUBJID",
#'   ylab = "SGOT/ALT (U/L)",
#'   add_baseline_hline = TRUE
#' )
#' p
#'
#' @export
h_g_ipp <- function(df,
                    xvar,
                    yvar,
                    xlab,
                    ylab,
                    id_var,
                    title = "Individual Patient Plots",
                    subtitle = "",
                    caption = NULL,
                    add_baseline_hline = FALSE,
                    yvar_baseline = "BASE",
                    ggtheme = nestcolor::theme_nest(),
                    col = NULL) {
  checkmate::assert_string(xvar)
  checkmate::assert_string(yvar)
  checkmate::assert_string(yvar_baseline)
  checkmate::assert_string(id_var)
  checkmate::assert_string(xlab)
  checkmate::assert_string(ylab)
  checkmate::assert_string(title)
  checkmate::assert_string(subtitle)
  checkmate::assert_subset(c(xvar, yvar, yvar_baseline, id_var), colnames(df))
  checkmate::assert_data_frame(df)
  checkmate::assert_flag(add_baseline_hline)
  checkmate::assert_character(col, null.ok = TRUE)

  p <- ggplot2::ggplot(
    data = df,
    mapping = ggplot2::aes(
      x = .data[[xvar]],
      y = .data[[yvar]],
      group = .data[[id_var]],
      colour = .data[[id_var]]
    )
  ) +
    ggplot2::geom_line(linewidth = 0.4) +
    ggplot2::geom_point(size = 2) +
    ggplot2::labs(
      x = xlab,
      y = ylab,
      title = title,
      subtitle = subtitle,
      caption = caption
    ) +
    ggtheme

  if (add_baseline_hline) {
    baseline_df <- df[, c(id_var, yvar_baseline)]
    baseline_df <- unique(baseline_df)

    p <- p +
      ggplot2::geom_hline(
        data = baseline_df,
        mapping = ggplot2::aes(
          yintercept = .data[[yvar_baseline]],
          colour = .data[[id_var]]
        ),
        linetype = "dotdash",
        linewidth = 0.4
      ) +
      ggplot2::geom_text(
        data = baseline_df,
        mapping = ggplot2::aes(
          x = 1,
          y = .data[[yvar_baseline]],
          label = .data[[id_var]],
          colour = .data[[id_var]]
        ),
        nudge_y = 0.025 * (max(df[, yvar], na.rm = TRUE) - min(df[, yvar], na.rm = TRUE)),
        vjust = "right",
        size = 2
      )

    if (!is.null(col)) {
      p <- p +
        ggplot2::scale_color_manual(values = col)
    }
  }
  p
}

#' @describeIn g_ipp Plotting function for individual patient plots which, depending on user
#'   preference, renders a single graphic or compiles a list of graphics that show trends in individual's parameter
#'   values over time.
#'
#' @return A `ggplot` object or a list of `ggplot` objects.
#'
#' @examples
#' library(dplyr)
#' library(nestcolor)
#'
#' # Select a small sample of data to plot.
#' adlb <- tern_ex_adlb %>%
#'   filter(PARAMCD == "ALT", !(AVISIT %in% c("SCREENING", "BASELINE"))) %>%
#'   slice(1:36)
#'
#' plot_list <- g_ipp(
#'   df = adlb,
#'   xvar = "AVISIT",
#'   yvar = "AVAL",
#'   xlab = "Visit",
#'   ylab = "SGOT/ALT (U/L)",
#'   title = "Individual Patient Plots",
#'   add_baseline_hline = TRUE,
#'   plotting_choices = "split_by_max_obs",
#'   max_obs_per_plot = 5
#' )
#' plot_list
#'
#' @export
g_ipp <- function(df,
                  xvar,
                  yvar,
                  xlab,
                  ylab,
                  id_var = "USUBJID",
                  title = "Individual Patient Plots",
                  subtitle = "",
                  caption = NULL,
                  add_baseline_hline = FALSE,
                  yvar_baseline = "BASE",
                  ggtheme = nestcolor::theme_nest(),
                  plotting_choices = c("all_in_one", "split_by_max_obs", "separate_by_obs"),
                  max_obs_per_plot = 4,
                  col = NULL) {
  checkmate::assert_count(max_obs_per_plot)
  checkmate::assert_subset(plotting_choices, c("all_in_one", "split_by_max_obs", "separate_by_obs"))
  checkmate::assert_character(col, null.ok = TRUE)

  plotting_choices <- match.arg(plotting_choices)

  if (plotting_choices == "all_in_one") {
    p <- h_g_ipp(
      df = df,
      xvar = xvar,
      yvar = yvar,
      xlab = xlab,
      ylab = ylab,
      id_var = id_var,
      title = title,
      subtitle = subtitle,
      caption = caption,
      add_baseline_hline = add_baseline_hline,
      yvar_baseline = yvar_baseline,
      ggtheme = ggtheme,
      col = col
    )

    return(p)
  } else if (plotting_choices == "split_by_max_obs") {
    id_vec <- unique(df[[id_var]])
    id_list <- split(
      id_vec,
      rep(1:ceiling(length(id_vec) / max_obs_per_plot),
        each = max_obs_per_plot,
        length.out = length(id_vec)
      )
    )

    df_list <- list()
    plot_list <- list()

    for (i in seq_along(id_list)) {
      df_list[[i]] <- df[df[[id_var]] %in% id_list[[i]], ]

      plots <- h_g_ipp(
        df = df_list[[i]],
        xvar = xvar,
        yvar = yvar,
        xlab = xlab,
        ylab = ylab,
        id_var = id_var,
        title = title,
        subtitle = subtitle,
        caption = caption,
        add_baseline_hline = add_baseline_hline,
        yvar_baseline = yvar_baseline,
        ggtheme = ggtheme,
        col = col
      )

      plot_list[[i]] <- plots
    }
    return(plot_list)
  } else {
    ind_df <- split(df, df[[id_var]])
    plot_list <- lapply(
      ind_df,
      function(x) {
        h_g_ipp(
          df = x,
          xvar = xvar,
          yvar = yvar,
          xlab = xlab,
          ylab = ylab,
          id_var = id_var,
          title = title,
          subtitle = subtitle,
          caption = caption,
          add_baseline_hline = add_baseline_hline,
          yvar_baseline = yvar_baseline,
          ggtheme = ggtheme,
          col = col
        )
      }
    )

    return(plot_list)
  }
}

#' Control functions for Kaplan-Meier plot annotation tables
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Auxiliary functions for controlling arguments for formatting the annotation tables that can be added to plots
#' generated via [g_km()].
#'
#' @param x (`proportion`)\cr x-coordinate for center of annotation table.
#' @param y (`proportion`)\cr y-coordinate for center of annotation table.
#' @param w (`proportion`)\cr relative width of the annotation table.
#' @param h (`proportion`)\cr relative height of the annotation table.
#' @param fill (`flag` or `character`)\cr whether the annotation table should have a background fill color.
#'   Can also be a color code to use as the background fill color. If `TRUE`, color code defaults to `"#00000020"`.
#'
#' @return A list of components with the same names as the arguments.
#'
#' @seealso [g_km()]
#'
#' @name control_annot
NULL

#' @describeIn control_annot Control function for formatting the median survival time annotation table. This annotation
#'   table can be added in [g_km()] by setting `annot_surv_med=TRUE`, and can be configured using the
#'   `control_surv_med_annot()` function by setting it as the `control_annot_surv_med` argument.
#'
#' @examples
#' control_surv_med_annot()
#'
#' @export
control_surv_med_annot <- function(x = 0.8, y = 0.85, w = 0.32, h = 0.16, fill = TRUE) {
  assert_proportion_value(x)
  assert_proportion_value(y)
  assert_proportion_value(w)
  assert_proportion_value(h)

  list(x = x, y = y, w = w, h = h, fill = fill)
}

#' @describeIn control_annot Control function for formatting the Cox-PH annotation table. This annotation table can be
#'   added in [g_km()] by setting `annot_coxph=TRUE`, and can be configured using the `control_coxph_annot()` function
#'   by setting it as the `control_annot_coxph` argument.
#'
#' @param ref_lbls (`flag`)\cr whether the reference group should be explicitly printed in labels for the
#'   annotation table. If `FALSE` (default), only comparison groups will be printed in the table labels.
#'
#' @examples
#' control_coxph_annot()
#'
#' @export
control_coxph_annot <- function(x = 0.29, y = 0.51, w = 0.4, h = 0.125, fill = TRUE, ref_lbls = FALSE) {
  checkmate::assert_logical(ref_lbls, any.missing = FALSE)

  res <- c(control_surv_med_annot(x = x, y = y, w = w, h = h), list(ref_lbls = ref_lbls))
  res
}

#' Helper function to calculate x-tick positions
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Calculate the positions of ticks on the x-axis. However, if `xticks` already
#' exists it is kept as is. It is based on the same function `ggplot2` relies on,
#' and is required in the graphic and the patient-at-risk annotation table.
#'
#' @inheritParams g_km
#' @inheritParams h_ggkm
#'
#' @return A vector of positions to use for x-axis ticks on a `ggplot` object.
#'
#' @examples
#' library(dplyr)
#' library(survival)
#'
#' data <- tern_ex_adtte %>%
#'   filter(PARAMCD == "OS") %>%
#'   survfit(formula = Surv(AVAL, 1 - CNSR) ~ ARMCD, data = .) %>%
#'   h_data_plot()
#'
#' h_xticks(data)
#' h_xticks(data, xticks = seq(0, 3000, 500))
#' h_xticks(data, xticks = 500)
#' h_xticks(data, xticks = 500, max_time = 6000)
#' h_xticks(data, xticks = c(0, 500), max_time = 300)
#' h_xticks(data, xticks = 500, max_time = 300)
#'
#' @export
h_xticks <- function(data, xticks = NULL, max_time = NULL) {
  if (is.null(xticks)) {
    if (is.null(max_time)) {
      labeling::extended(range(data$time)[1], range(data$time)[2], m = 5)
    } else {
      labeling::extended(range(data$time)[1], max(range(data$time)[2], max_time), m = 5)
    }
  } else if (checkmate::test_number(xticks)) {
    if (is.null(max_time)) {
      seq(0, max(data$time), xticks)
    } else {
      seq(0, max(data$time, max_time), xticks)
    }
  } else if (is.numeric(xticks)) {
    xticks
  } else {
    stop(
      paste(
        "xticks should be either `NULL`",
        "or a single number (interval between x ticks)",
        "or a numeric vector (position of ticks on the x axis)"
      )
    )
  }
}

#' Helper function for survival estimations
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Transform a survival fit to a table with groups in rows characterized by N, median and confidence interval.
#'
#' @inheritParams h_data_plot
#'
#' @return A summary table with statistics `N`, `Median`, and `XX% CI` (`XX` taken from `fit_km`).
#'
#' @examples
#' library(dplyr)
#' library(survival)
#'
#' adtte <- tern_ex_adtte %>% filter(PARAMCD == "OS")
#' fit <- survfit(
#'   formula = Surv(AVAL, 1 - CNSR) ~ ARMCD,
#'   data = adtte
#' )
#' h_tbl_median_surv(fit_km = fit)
#'
#' @export
h_tbl_median_surv <- function(fit_km, armval = "All") {
  y <- if (is.null(fit_km$strata)) {
    as.data.frame(t(summary(fit_km)$table), row.names = armval)
  } else {
    tbl <- summary(fit_km)$table
    rownames_lst <- strsplit(sub("=", "equals", rownames(tbl)), "equals")
    rownames(tbl) <- matrix(unlist(rownames_lst), ncol = 2, byrow = TRUE)[, 2]
    as.data.frame(tbl)
  }
  conf.int <- summary(fit_km)$conf.int # nolint
  y$records <- round(y$records)
  y$median <- signif(y$median, 4)
  y$`CI` <- paste0(
    "(", signif(y[[paste0(conf.int, "LCL")]], 4), ", ", signif(y[[paste0(conf.int, "UCL")]], 4), ")"
  )
  stats::setNames(
    y[c("records", "median", "CI")],
    c("N", "Median", f_conf_level(conf.int))
  )
}

#' Helper function for generating a pairwise Cox-PH table
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Create a `data.frame` of pairwise stratified or unstratified Cox-PH analysis results.
#'
#' @inheritParams g_km
#' @param annot_coxph_ref_lbls (`flag`)\cr whether the reference group should be explicitly printed in labels for the
#'   `annot_coxph` table. If `FALSE` (default), only comparison groups will be printed in `annot_coxph` table labels.
#'
#' @return A `data.frame` containing statistics `HR`, `XX% CI` (`XX` taken from `control_coxph_pw`),
#'   and `p-value (log-rank)`.
#'
#' @examples
#' library(dplyr)
#'
#' adtte <- tern_ex_adtte %>%
#'   filter(PARAMCD == "OS") %>%
#'   mutate(is_event = CNSR == 0)
#'
#' h_tbl_coxph_pairwise(
#'   df = adtte,
#'   variables = list(tte = "AVAL", is_event = "is_event", arm = "ARM"),
#'   control_coxph_pw = control_coxph(conf_level = 0.9)
#' )
#'
#' @export
h_tbl_coxph_pairwise <- function(df,
                                 variables,
                                 ref_group_coxph = NULL,
                                 control_coxph_pw = control_coxph(),
                                 annot_coxph_ref_lbls = FALSE) {
  if ("strat" %in% names(variables)) {
    warning(
      "Warning: the `strat` element name of the `variables` list argument to `h_tbl_coxph_pairwise() ",
      "was deprecated in tern 0.9.4.\n  ",
      "Please use the name `strata` instead of `strat` in the `variables` argument."
    )
    variables[["strata"]] <- variables[["strat"]]
  }

  assert_df_with_variables(df, variables)
  checkmate::assert_choice(ref_group_coxph, levels(df[[variables$arm]]), null.ok = TRUE)
  checkmate::assert_flag(annot_coxph_ref_lbls)

  arm <- variables$arm
  df[[arm]] <- factor(df[[arm]])

  ref_group <- if (!is.null(ref_group_coxph)) ref_group_coxph else levels(df[[variables$arm]])[1]
  comp_group <- setdiff(levels(df[[arm]]), ref_group)

  results <- Map(function(comp) {
    res <- s_coxph_pairwise(
      df = df[df[[arm]] == comp, , drop = FALSE],
      .ref_group = df[df[[arm]] == ref_group, , drop = FALSE],
      .in_ref_col = FALSE,
      .var = variables$tte,
      is_event = variables$is_event,
      strata = variables$strata,
      control = control_coxph_pw
    )
    res_df <- data.frame(
      hr = format(round(res$hr, 2), nsmall = 2),
      hr_ci = paste0(
        "(", format(round(res$hr_ci[1], 2), nsmall = 2), ", ",
        format(round(res$hr_ci[2], 2), nsmall = 2), ")"
      ),
      pvalue = if (res$pvalue < 0.0001) "<0.0001" else format(round(res$pvalue, 4), 4),
      stringsAsFactors = FALSE
    )
    colnames(res_df) <- c("HR", vapply(res[c("hr_ci", "pvalue")], obj_label, FUN.VALUE = "character"))
    row.names(res_df) <- comp
    res_df
  }, comp_group)
  if (annot_coxph_ref_lbls) names(results) <- paste(comp_group, "vs.", ref_group)

  do.call(rbind, results)
}

#' Helper function to tidy survival fit data
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Convert the survival fit data into a data frame designed for plotting
#' within `g_km`.
#'
#' This starts from the [broom::tidy()] result, and then:
#'   * Post-processes the `strata` column into a factor.
#'   * Extends each stratum by an additional first row with time 0 and probability 1 so that
#'     downstream plot lines start at those coordinates.
#'   * Adds a `censor` column.
#'   * Filters the rows before `max_time`.
#'
#' @inheritParams g_km
#' @param fit_km (`survfit`)\cr result of [survival::survfit()].
#' @param armval (`string`)\cr used as strata name when treatment arm variable only has one level. Default is `"All"`.
#'
#' @return A `tibble` with columns `time`, `n.risk`, `n.event`, `n.censor`, `estimate`, `std.error`, `conf.high`,
#'   `conf.low`, `strata`, and `censor`.
#'
#' @examples
#' library(dplyr)
#' library(survival)
#'
#' # Test with multiple arms
#' tern_ex_adtte %>%
#'   filter(PARAMCD == "OS") %>%
#'   survfit(formula = Surv(AVAL, 1 - CNSR) ~ ARMCD, data = .) %>%
#'   h_data_plot()
#'
#' # Test with single arm
#' tern_ex_adtte %>%
#'   filter(PARAMCD == "OS", ARMCD == "ARM B") %>%
#'   survfit(formula = Surv(AVAL, 1 - CNSR) ~ ARMCD, data = .) %>%
#'   h_data_plot(armval = "ARM B")
#'
#' @export
h_data_plot <- function(fit_km,
                        armval = "All",
                        max_time = NULL) {
  y <- broom::tidy(fit_km)

  if (!is.null(fit_km$strata)) {
    fit_km_var_level <- strsplit(sub("=", "equals", names(fit_km$strata)), "equals")
    strata_levels <- vapply(fit_km_var_level, FUN = "[", FUN.VALUE = "a", i = 2)
    strata_var_level <- strsplit(sub("=", "equals", y$strata), "equals")
    y$strata <- factor(
      vapply(strata_var_level, FUN = "[", FUN.VALUE = "a", i = 2),
      levels = strata_levels
    )
  } else {
    y$strata <- armval
  }

  y_by_strata <- split(y, y$strata)
  y_by_strata_extended <- lapply(
    y_by_strata,
    FUN = function(tbl) {
      first_row <- tbl[1L, ]
      first_row$time <- 0
      first_row$n.risk <- sum(first_row[, c("n.risk", "n.event", "n.censor")])
      first_row$n.event <- first_row$n.censor <- 0
      first_row$estimate <- first_row$conf.high <- first_row$conf.low <- 1
      first_row$std.error <- 0
      rbind(
        first_row,
        tbl
      )
    }
  )
  y <- do.call(rbind, y_by_strata_extended)

  y$censor <- ifelse(y$n.censor > 0, y$estimate, NA)
  if (!is.null(max_time)) {
    y <- y[y$time <= max(max_time), ]
  }
  y
}

## Deprecated Functions ----

#' Helper function to create a KM plot
#'
#' @description `r lifecycle::badge("deprecated")`
#'
#' Draw the Kaplan-Meier plot using `ggplot2`.
#'
#' @inheritParams g_km
#' @param data (`data.frame`)\cr survival data as pre-processed by `h_data_plot`.
#'
#' @return A `ggplot` object.
#'
#' @examples
#' \donttest{
#' library(dplyr)
#' library(survival)
#'
#' fit_km <- tern_ex_adtte %>%
#'   filter(PARAMCD == "OS") %>%
#'   survfit(formula = Surv(AVAL, 1 - CNSR) ~ ARMCD, data = .)
#' data_plot <- h_data_plot(fit_km = fit_km)
#' xticks <- h_xticks(data = data_plot)
#' gg <- h_ggkm(
#'   data = data_plot,
#'   censor_show = TRUE,
#'   xticks = xticks,
#'   xlab = "Days",
#'   yval = "Survival",
#'   ylab = "Survival Probability",
#'   title = "Survival"
#' )
#' gg
#' }
#'
#' @export
h_ggkm <- function(data,
                   xticks = NULL,
                   yval = "Survival",
                   censor_show,
                   xlab,
                   ylab,
                   ylim = NULL,
                   title,
                   footnotes = NULL,
                   max_time = NULL,
                   lwd = 1,
                   lty = NULL,
                   pch = 3,
                   size = 2,
                   col = NULL,
                   ci_ribbon = FALSE,
                   ggtheme = nestcolor::theme_nest()) {
  lifecycle::deprecate_warn(
    "0.9.4",
    "h_ggkm()",
    details = "`g_km` now generates `ggplot` objects. This function is no longer used within `tern`."
  )
  checkmate::assert_numeric(lty, null.ok = TRUE)
  checkmate::assert_character(col, null.ok = TRUE)

  if (is.null(ylim)) {
    data_lims <- data
    if (yval == "Failure") data_lims[["estimate"]] <- 1 - data_lims[["estimate"]]
    if (!is.null(max_time)) {
      y_lwr <- min(data_lims[data_lims$time < max_time, ][["estimate"]])
      y_upr <- max(data_lims[data_lims$time < max_time, ][["estimate"]])
    } else {
      y_lwr <- min(data_lims[["estimate"]])
      y_upr <- max(data_lims[["estimate"]])
    }
    ylim <- c(y_lwr, y_upr)
  }
  checkmate::assert_numeric(ylim, finite = TRUE, any.missing = FALSE, len = 2, sorted = TRUE)

  # change estimates of survival to estimates of failure (1 - survival)
  if (yval == "Failure") {
    data$estimate <- 1 - data$estimate
    data[c("conf.high", "conf.low")] <- list(1 - data$conf.low, 1 - data$conf.high)
    data$censor <- 1 - data$censor
  }

  gg <- {
    ggplot2::ggplot(
      data = data,
      mapping = ggplot2::aes(
        x = .data[["time"]],
        y = .data[["estimate"]],
        ymin = .data[["conf.low"]],
        ymax = .data[["conf.high"]],
        color = .data[["strata"]],
        fill = .data[["strata"]]
      )
    ) +
      ggplot2::geom_hline(yintercept = 0)
  }

  if (ci_ribbon) {
    gg <- gg + ggplot2::geom_ribbon(alpha = .3, lty = 0)
  }

  gg <- if (is.null(lty)) {
    gg +
      ggplot2::geom_step(linewidth = lwd)
  } else if (checkmate::test_number(lty)) {
    gg +
      ggplot2::geom_step(linewidth = lwd, lty = lty)
  } else if (is.numeric(lty)) {
    gg +
      ggplot2::geom_step(mapping = ggplot2::aes(linetype = .data[["strata"]]), linewidth = lwd) +
      ggplot2::scale_linetype_manual(values = lty)
  }

  gg <- gg +
    ggplot2::coord_cartesian(ylim = ylim) +
    ggplot2::labs(x = xlab, y = ylab, title = title, caption = footnotes)

  if (!is.null(col)) {
    gg <- gg +
      ggplot2::scale_color_manual(values = col) +
      ggplot2::scale_fill_manual(values = col)
  }
  if (censor_show) {
    dt <- data[data$n.censor != 0, ]
    dt$censor_lbl <- factor("Censored")

    gg <- gg + ggplot2::geom_point(
      data = dt,
      ggplot2::aes(
        x = .data[["time"]],
        y = .data[["censor"]],
        shape = .data[["censor_lbl"]]
      ),
      size = size,
      show.legend = TRUE,
      inherit.aes = TRUE
    ) +
      ggplot2::scale_shape_manual(name = NULL, values = pch) +
      ggplot2::guides(
        shape = ggplot2::guide_legend(override.aes = list(linetype = NA)),
        fill = ggplot2::guide_legend(override.aes = list(shape = NA))
      )
  }

  if (!is.null(max_time) && !is.null(xticks)) {
    gg <- gg + ggplot2::scale_x_continuous(breaks = xticks, limits = c(min(0, xticks), max(c(xticks, max_time))))
  } else if (!is.null(xticks)) {
    if (max(data$time) <= max(xticks)) {
      gg <- gg + ggplot2::scale_x_continuous(breaks = xticks, limits = c(min(0, min(xticks)), max(xticks)))
    } else {
      gg <- gg + ggplot2::scale_x_continuous(breaks = xticks)
    }
  } else if (!is.null(max_time)) {
    gg <- gg + ggplot2::scale_x_continuous(limits = c(0, max_time))
  }

  if (!is.null(ggtheme)) {
    gg <- gg + ggtheme
  }

  gg + ggplot2::theme(
    legend.position = "bottom",
    legend.title = ggplot2::element_blank(),
    legend.key.height = unit(0.02, "npc"),
    panel.grid.major.x = ggplot2::element_line(linewidth = 2)
  )
}

#' `ggplot` decomposition
#'
#' @description `r lifecycle::badge("deprecated")`
#'
#' The elements composing the `ggplot` are extracted and organized in a `list`.
#'
#' @param gg (`ggplot`)\cr a graphic to decompose.
#'
#' @return A named `list` with elements:
#'   * `panel`: The panel.
#'   * `yaxis`: The y-axis.
#'   * `xaxis`: The x-axis.
#'   * `xlab`: The x-axis label.
#'   * `ylab`: The y-axis label.
#'   * `guide`: The legend.
#'
#' @examples
#' \donttest{
#' library(dplyr)
#' library(survival)
#' library(grid)
#'
#' fit_km <- tern_ex_adtte %>%
#'   filter(PARAMCD == "OS") %>%
#'   survfit(formula = Surv(AVAL, 1 - CNSR) ~ ARMCD, data = .)
#' data_plot <- h_data_plot(fit_km = fit_km)
#' xticks <- h_xticks(data = data_plot)
#' gg <- h_ggkm(
#'   data = data_plot,
#'   yval = "Survival",
#'   censor_show = TRUE,
#'   xticks = xticks, xlab = "Days", ylab = "Survival Probability",
#'   title = "tt",
#'   footnotes = "ff"
#' )
#'
#' g_el <- h_decompose_gg(gg)
#' grid::grid.newpage()
#' grid.rect(gp = grid::gpar(lty = 1, col = "red", fill = "gray85", lwd = 5))
#' grid::grid.draw(g_el$panel)
#'
#' grid::grid.newpage()
#' grid.rect(gp = grid::gpar(lty = 1, col = "royalblue", fill = "gray85", lwd = 5))
#' grid::grid.draw(with(g_el, cbind(ylab, yaxis)))
#' }
#'
#' @export
h_decompose_gg <- function(gg) {
  lifecycle::deprecate_warn(
    "0.9.4",
    "h_decompose_gg()",
    details = "`g_km` now generates `ggplot` objects. This function is no longer used within `tern`."
  )
  g_el <- ggplot2::ggplotGrob(gg)
  y <- c(
    panel = "panel",
    yaxis = "axis-l",
    xaxis = "axis-b",
    xlab = "xlab-b",
    ylab = "ylab-l",
    guide = "guide"
  )
  lapply(X = y, function(x) gtable::gtable_filter(g_el, x))
}

#' Helper function to prepare a KM layout
#'
#' @description `r lifecycle::badge("deprecated")`
#'
#' Prepares a (5 rows) x (2 cols) layout for the Kaplan-Meier curve.
#'
#' @inheritParams g_km
#' @inheritParams h_ggkm
#' @param g_el (`list` of `gtable`)\cr list as obtained by `h_decompose_gg()`.
#' @param annot_at_risk (`flag`)\cr compute and add the annotation table reporting the number of
#'   patient at risk matching the main grid of the Kaplan-Meier curve.
#'
#' @return A grid layout.
#'
#' @details The layout corresponds to a grid of two columns and five rows of unequal dimensions. Most of the
#'   dimension are fixed, only the curve is flexible and will accommodate with the remaining free space.
#'   * The left column gets the annotation of the `ggplot` (y-axis) and the names of the strata for the patient
#'     at risk tabulation. The main constraint is about the width of the columns which must allow the writing of
#'     the strata name.
#'   * The right column receive the `ggplot`, the legend, the x-axis and the patient at risk table.
#'
#' @examples
#' \donttest{
#' library(dplyr)
#' library(survival)
#' library(grid)
#'
#' fit_km <- tern_ex_adtte %>%
#'   filter(PARAMCD == "OS") %>%
#'   survfit(formula = Surv(AVAL, 1 - CNSR) ~ ARMCD, data = .)
#' data_plot <- h_data_plot(fit_km = fit_km)
#' xticks <- h_xticks(data = data_plot)
#' gg <- h_ggkm(
#'   data = data_plot,
#'   censor_show = TRUE,
#'   xticks = xticks, xlab = "Days", ylab = "Survival Probability",
#'   title = "tt", footnotes = "ff", yval = "Survival"
#' )
#' g_el <- h_decompose_gg(gg)
#' lyt <- h_km_layout(data = data_plot, g_el = g_el, title = "t", footnotes = "f")
#' grid.show.layout(lyt)
#' }
#'
#' @export
h_km_layout <- function(data, g_el, title, footnotes, annot_at_risk = TRUE, annot_at_risk_title = TRUE) {
  lifecycle::deprecate_warn(
    "0.9.4",
    "h_km_layout()",
    details = "`g_km` now generates `ggplot` objects. This function is no longer used within `tern`."
  )
  txtlines <- levels(as.factor(data$strata))
  nlines <- nlevels(as.factor(data$strata))
  col_annot_width <- max(
    c(
      as.numeric(grid::convertX(g_el$yaxis$widths + g_el$ylab$widths, "pt")),
      as.numeric(
        grid::convertX(
          grid::stringWidth(txtlines) + grid::unit(7, "pt"), "pt"
        )
      )
    )
  )

  ttl_row <- as.numeric(!is.null(title))
  foot_row <- as.numeric(!is.null(footnotes))
  no_tbl_ind <- c()
  ht_x <- c()
  ht_units <- c()

  if (ttl_row == 1) {
    no_tbl_ind <- c(no_tbl_ind, TRUE)
    ht_x <- c(ht_x, 2)
    ht_units <- c(ht_units, "lines")
  }

  no_tbl_ind <- c(no_tbl_ind, rep(TRUE, 3), rep(FALSE, 2))
  ht_x <- c(
    ht_x,
    1,
    grid::convertX(with(g_el, xaxis$heights + ylab$widths), "pt") + grid::unit(5, "pt"),
    grid::convertX(g_el$guide$heights, "pt") + grid::unit(2, "pt"),
    1,
    nlines + 0.5,
    grid::convertX(with(g_el, xaxis$heights + ylab$widths), "pt")
  )
  ht_units <- c(
    ht_units,
    "null",
    "pt",
    "pt",
    "lines",
    "lines",
    "pt"
  )

  if (foot_row == 1) {
    no_tbl_ind <- c(no_tbl_ind, TRUE)
    ht_x <- c(ht_x, 1)
    ht_units <- c(ht_units, "lines")
  }
  if (annot_at_risk) {
    no_at_risk_tbl <- rep(TRUE, 6 + ttl_row + foot_row)
    if (!annot_at_risk_title) {
      no_at_risk_tbl[length(no_at_risk_tbl) - 2 - foot_row] <- FALSE
    }
  } else {
    no_at_risk_tbl <- no_tbl_ind
  }

  grid::grid.layout(
    nrow = sum(no_at_risk_tbl), ncol = 2,
    widths = grid::unit(c(col_annot_width, 1), c("pt", "null")),
    heights = grid::unit(
      x = ht_x[no_at_risk_tbl],
      units = ht_units[no_at_risk_tbl]
    )
  )
}

#' Helper function to create patient-at-risk grobs
#'
#' @description `r lifecycle::badge("deprecated")`
#'
#' Two graphical objects are obtained, one corresponding to row labeling and the second to the table of
#' numbers of patients at risk. If `title = TRUE`, a third object corresponding to the table title is
#' also obtained.
#'
#' @inheritParams g_km
#' @inheritParams h_ggkm
#' @param annot_tbl (`data.frame`)\cr annotation as prepared by [survival::summary.survfit()] which
#'   includes the number of patients at risk at given time points.
#' @param xlim (`numeric(1)`)\cr the maximum value on the x-axis (used to ensure the at risk table aligns with the KM
#'   graph).
#' @param title (`flag`)\cr whether the "Patients at Risk" title should be added above the `annot_at_risk`
#'   table. Has no effect if `annot_at_risk` is `FALSE`. Defaults to `TRUE`.
#'
#' @return A named `list` of two `gTree` objects if `title = FALSE`: `at_risk` and `label`, or three
#'   `gTree` objects if `title = TRUE`: `at_risk`, `label`, and `title`.
#'
#' @examples
#' \donttest{
#' library(dplyr)
#' library(survival)
#' library(grid)
#'
#' fit_km <- tern_ex_adtte %>%
#'   filter(PARAMCD == "OS") %>%
#'   survfit(formula = Surv(AVAL, 1 - CNSR) ~ ARMCD, data = .)
#'
#' data_plot <- h_data_plot(fit_km = fit_km)
#'
#' xticks <- h_xticks(data = data_plot)
#'
#' gg <- h_ggkm(
#'   data = data_plot,
#'   censor_show = TRUE,
#'   xticks = xticks, xlab = "Days", ylab = "Survival Probability",
#'   title = "tt", footnotes = "ff", yval = "Survival"
#' )
#'
#' # The annotation table reports the patient at risk for a given strata and
#' # times (`xticks`).
#' annot_tbl <- summary(fit_km, times = xticks)
#' if (is.null(fit_km$strata)) {
#'   annot_tbl <- with(annot_tbl, data.frame(n.risk = n.risk, time = time, strata = "All"))
#' } else {
#'   strata_lst <- strsplit(sub("=", "equals", levels(annot_tbl$strata)), "equals")
#'   levels(annot_tbl$strata) <- matrix(unlist(strata_lst), ncol = 2, byrow = TRUE)[, 2]
#'   annot_tbl <- data.frame(
#'     n.risk = annot_tbl$n.risk,
#'     time = annot_tbl$time,
#'     strata = annot_tbl$strata
#'   )
#' }
#'
#' # The annotation table is transformed into a grob.
#' tbl <- h_grob_tbl_at_risk(data = data_plot, annot_tbl = annot_tbl, xlim = max(xticks))
#'
#' # For the representation, the layout is estimated for which the decomposition
#' # of the graphic element is necessary.
#' g_el <- h_decompose_gg(gg)
#' lyt <- h_km_layout(data = data_plot, g_el = g_el, title = "t", footnotes = "f")
#'
#' grid::grid.newpage()
#' pushViewport(viewport(layout = lyt, height = .95, width = .95))
#' grid.rect(gp = grid::gpar(lty = 1, col = "purple", fill = "gray85", lwd = 1))
#' pushViewport(viewport(layout.pos.row = 3:4, layout.pos.col = 2))
#' grid.rect(gp = grid::gpar(lty = 1, col = "orange", fill = "gray85", lwd = 1))
#' grid::grid.draw(tbl$at_risk)
#' popViewport()
#' pushViewport(viewport(layout.pos.row = 3:4, layout.pos.col = 1))
#' grid.rect(gp = grid::gpar(lty = 1, col = "green3", fill = "gray85", lwd = 1))
#' grid::grid.draw(tbl$label)
#' }
#'
#' @export
h_grob_tbl_at_risk <- function(data, annot_tbl, xlim, title = TRUE) {
  lifecycle::deprecate_warn(
    "0.9.4",
    "h_grob_tbl_at_risk()",
    details = "`g_km` now generates `ggplot` objects. This function is no longer used within `tern`."
  )
  txtlines <- levels(as.factor(data$strata))
  nlines <- nlevels(as.factor(data$strata))
  y_int <- annot_tbl$time[2] - annot_tbl$time[1]
  annot_tbl <- expand.grid(
    time = seq(0, xlim, y_int),
    strata = unique(annot_tbl$strata)
  ) %>% dplyr::left_join(annot_tbl, by = c("time", "strata"))
  annot_tbl[is.na(annot_tbl)] <- 0
  y_str_unit <- as.numeric(annot_tbl$strata)
  vp_table <- grid::plotViewport(margins = grid::unit(c(0, 0, 0, 0), "lines"))
  if (title) {
    gb_table_title <- grid::gList(
      grid::textGrob(
        label = "Patients at Risk:",
        x = 1,
        y = grid::unit(0.2, "native"),
        gp = grid::gpar(fontface = "bold", fontsize = 10)
      )
    )
  }
  gb_table_left_annot <- grid::gList(
    grid::rectGrob(
      x = 0, y = grid::unit(c(1:nlines) - 1, "lines"),
      gp = grid::gpar(fill = c("gray95", "gray90"), alpha = 1, col = "white"),
      height = grid::unit(1, "lines"), just = "bottom", hjust = 0
    ),
    grid::textGrob(
      label = unique(annot_tbl$strata),
      x = 0.5,
      y = grid::unit(
        (max(unique(y_str_unit)) - unique(y_str_unit)) + 0.75,
        "native"
      ),
      gp = grid::gpar(fontface = "italic", fontsize = 10)
    )
  )
  gb_patient_at_risk <- grid::gList(
    grid::rectGrob(
      x = 0, y = grid::unit(c(1:nlines) - 1, "lines"),
      gp = grid::gpar(fill = c("gray95", "gray90"), alpha = 1, col = "white"),
      height = grid::unit(1, "lines"), just = "bottom", hjust = 0
    ),
    grid::textGrob(
      label = annot_tbl$n.risk,
      x = grid::unit(annot_tbl$time, "native"),
      y = grid::unit(
        (max(y_str_unit) - y_str_unit) + .5,
        "line"
      ) # maybe native
    )
  )

  ret <- list(
    at_risk = grid::gList(
      grid::gTree(
        vp = vp_table,
        children = grid::gList(
          grid::gTree(
            vp = grid::dataViewport(
              xscale = c(0, xlim) + c(-0.05, 0.05) * xlim,
              yscale = c(0, nlines + 1),
              extension = c(0.05, 0)
            ),
            children = grid::gList(gb_patient_at_risk)
          )
        )
      )
    ),
    label = grid::gList(
      grid::gTree(
        vp = grid::viewport(width = max(grid::stringWidth(txtlines))),
        children = grid::gList(
          grid::gTree(
            vp = grid::dataViewport(
              xscale = 0:1,
              yscale = c(0, nlines + 1),
              extension = c(0.0, 0)
            ),
            children = grid::gList(gb_table_left_annot)
          )
        )
      )
    )
  )

  if (title) {
    ret[["title"]] <- grid::gList(
      grid::gTree(
        vp = grid::viewport(width = max(grid::stringWidth(txtlines))),
        children = grid::gList(
          grid::gTree(
            vp = grid::dataViewport(
              xscale = 0:1,
              yscale = c(0, 1),
              extension = c(0, 0)
            ),
            children = grid::gList(gb_table_title)
          )
        )
      )
    )
  }

  ret
}

#' Helper function to create survival estimation grobs
#'
#' @description `r lifecycle::badge("deprecated")`
#'
#' The survival fit is transformed in a grob containing a table with groups in
#' rows characterized by N, median and 95% confidence interval.
#'
#' @inheritParams g_km
#' @inheritParams h_data_plot
#' @param ttheme (`list`)\cr see [gridExtra::ttheme_default()].
#' @param x (`proportion`)\cr a value between 0 and 1 specifying x-location.
#' @param y (`proportion`)\cr a value between 0 and 1 specifying y-location.
#' @param width (`grid::unit`)\cr width (as a unit) to use when printing the grob.
#'
#' @return A `grob` of a table containing statistics `N`, `Median`, and `XX% CI` (`XX` taken from `fit_km`).
#'
#' @examples
#' \donttest{
#' library(dplyr)
#' library(survival)
#' library(grid)
#'
#' grid::grid.newpage()
#' grid.rect(gp = grid::gpar(lty = 1, col = "pink", fill = "gray85", lwd = 1))
#' tern_ex_adtte %>%
#'   filter(PARAMCD == "OS") %>%
#'   survfit(formula = Surv(AVAL, 1 - CNSR) ~ ARMCD, data = .) %>%
#'   h_grob_median_surv() %>%
#'   grid::grid.draw()
#' }
#'
#' @export
h_grob_median_surv <- function(fit_km,
                               armval = "All",
                               x = 0.9,
                               y = 0.9,
                               width = grid::unit(0.3, "npc"),
                               ttheme = gridExtra::ttheme_default()) {
  lifecycle::deprecate_warn(
    "0.9.4",
    "h_grob_median_surv()",
    details = "`g_km` now generates `ggplot` objects. This function is no longer used within `tern`."
  )
  data <- h_tbl_median_surv(fit_km, armval = armval)

  width <- grid::convertUnit(grid::unit(as.numeric(width), grid::unitType(width)), "in")
  height <- width * (nrow(data) + 1) / 12

  w <- paste(" ", c(
    rownames(data)[which.max(nchar(rownames(data)))],
    sapply(names(data), function(x) c(x, data[[x]])[which.max(nchar(c(x, data[[x]])))])
  ))
  w_unit <- grid::convertWidth(grid::stringWidth(w), "in", valueOnly = TRUE)

  w_txt <- sapply(1:64, function(x) {
    graphics::par(ps = x)
    graphics::strwidth(w[4], units = "in")
  })
  f_size_w <- which.max(w_txt[w_txt < as.numeric((w_unit / sum(w_unit)) * width)[4]])

  h_txt <- sapply(1:64, function(x) {
    graphics::par(ps = x)
    graphics::strheight(grid::stringHeight("X"), units = "in")
  })
  f_size_h <- which.max(h_txt[h_txt < as.numeric(grid::unit(as.numeric(height) / 4, grid::unitType(height)))])

  if (ttheme$core$fg_params$fontsize == 12) {
    ttheme$core$fg_params$fontsize <- min(f_size_w, f_size_h)
    ttheme$colhead$fg_params$fontsize <- min(f_size_w, f_size_h)
    ttheme$rowhead$fg_params$fontsize <- min(f_size_w, f_size_h)
  }

  gt <- gridExtra::tableGrob(
    d = data,
    theme = ttheme
  )
  gt$widths <- ((w_unit / sum(w_unit)) * width)
  gt$heights <- rep(grid::unit(as.numeric(height) / 4, grid::unitType(height)), nrow(gt))

  vp <- grid::viewport(
    x = grid::unit(x, "npc") + grid::unit(1, "lines"),
    y = grid::unit(y, "npc") + grid::unit(1.5, "lines"),
    height = height,
    width = width,
    just = c("right", "top")
  )

  grid::gList(
    grid::gTree(
      vp = vp,
      children = grid::gList(gt)
    )
  )
}

#' Helper function to create grid object with y-axis annotation
#'
#' @description `r lifecycle::badge("deprecated")`
#'
#' Build the y-axis annotation from a decomposed `ggplot`.
#'
#' @param ylab (`gtable`)\cr the y-lab as a graphical object derived from a `ggplot`.
#' @param yaxis (`gtable`)\cr the y-axis as a graphical object derived from a `ggplot`.
#'
#' @return A `gTree` object containing the y-axis annotation from a `ggplot`.
#'
#' @examples
#' \donttest{
#' library(dplyr)
#' library(survival)
#' library(grid)
#'
#' fit_km <- tern_ex_adtte %>%
#'   filter(PARAMCD == "OS") %>%
#'   survfit(formula = Surv(AVAL, 1 - CNSR) ~ ARMCD, data = .)
#' data_plot <- h_data_plot(fit_km = fit_km)
#' xticks <- h_xticks(data = data_plot)
#' gg <- h_ggkm(
#'   data = data_plot,
#'   censor_show = TRUE,
#'   xticks = xticks, xlab = "Days", ylab = "Survival Probability",
#'   title = "title", footnotes = "footnotes", yval = "Survival"
#' )
#'
#' g_el <- h_decompose_gg(gg)
#'
#' grid::grid.newpage()
#' pvp <- grid::plotViewport(margins = c(5, 4, 2, 20))
#' pushViewport(pvp)
#' grid::grid.draw(h_grob_y_annot(ylab = g_el$ylab, yaxis = g_el$yaxis))
#' grid.rect(gp = grid::gpar(lty = 1, col = "gray35", fill = NA))
#' }
#'
#' @export
h_grob_y_annot <- function(ylab, yaxis) {
  lifecycle::deprecate_warn(
    "0.9.4",
    "h_grob_y_annot()",
    details = "`g_km` now generates `ggplot` objects. This function is no longer used within `tern`."
  )
  grid::gList(
    grid::gTree(
      vp = grid::viewport(
        width = grid::convertX(yaxis$widths + ylab$widths, "pt"),
        x = grid::unit(1, "npc"),
        just = "right"
      ),
      children = grid::gList(cbind(ylab, yaxis))
    )
  )
}

#' Helper function to create Cox-PH grobs
#'
#' @description `r lifecycle::badge("deprecated")`
#'
#' Grob of `rtable` output from [h_tbl_coxph_pairwise()]
#'
#' @inheritParams h_grob_median_surv
#' @param ... arguments to pass to [h_tbl_coxph_pairwise()].
#' @param x (`proportion`)\cr a value between 0 and 1 specifying x-location.
#' @param y (`proportion`)\cr a value between 0 and 1 specifying y-location.
#' @param width (`grid::unit`)\cr width (as a unit) to use when printing the grob.
#'
#' @return A `grob` of a table containing statistics `HR`, `XX% CI` (`XX` taken from `control_coxph_pw`),
#'   and `p-value (log-rank)`.
#'
#' @examples
#' \donttest{
#' library(dplyr)
#' library(survival)
#' library(grid)
#'
#' grid::grid.newpage()
#' grid.rect(gp = grid::gpar(lty = 1, col = "pink", fill = "gray85", lwd = 1))
#' data <- tern_ex_adtte %>%
#'   filter(PARAMCD == "OS") %>%
#'   mutate(is_event = CNSR == 0)
#' tbl_grob <- h_grob_coxph(
#'   df = data,
#'   variables = list(tte = "AVAL", is_event = "is_event", arm = "ARMCD"),
#'   control_coxph_pw = control_coxph(conf_level = 0.9), x = 0.5, y = 0.5
#' )
#' grid::grid.draw(tbl_grob)
#' }
#'
#' @export
h_grob_coxph <- function(...,
                         x = 0,
                         y = 0,
                         width = grid::unit(0.4, "npc"),
                         ttheme = gridExtra::ttheme_default(
                           padding = grid::unit(c(1, .5), "lines"),
                           core = list(bg_params = list(fill = c("grey95", "grey90"), alpha = .5))
                         )) {
  lifecycle::deprecate_warn(
    "0.9.4",
    "h_grob_coxph()",
    details = "`g_km` now generates `ggplot` objects. This function is no longer used within `tern`."
  )
  data <- h_tbl_coxph_pairwise(...)

  width <- grid::convertUnit(grid::unit(as.numeric(width), grid::unitType(width)), "in")
  height <- width * (nrow(data) + 1) / 12

  w <- paste("    ", c(
    rownames(data)[which.max(nchar(rownames(data)))],
    sapply(names(data), function(x) c(x, data[[x]])[which.max(nchar(c(x, data[[x]])))])
  ))
  w_unit <- grid::convertWidth(grid::stringWidth(w), "in", valueOnly = TRUE)

  w_txt <- sapply(1:64, function(x) {
    graphics::par(ps = x)
    graphics::strwidth(w[4], units = "in")
  })
  f_size_w <- which.max(w_txt[w_txt < as.numeric((w_unit / sum(w_unit)) * width)[4]])

  h_txt <- sapply(1:64, function(x) {
    graphics::par(ps = x)
    graphics::strheight(grid::stringHeight("X"), units = "in")
  })
  f_size_h <- which.max(h_txt[h_txt < as.numeric(grid::unit(as.numeric(height) / 4, grid::unitType(height)))])

  if (ttheme$core$fg_params$fontsize == 12) {
    ttheme$core$fg_params$fontsize <- min(f_size_w, f_size_h)
    ttheme$colhead$fg_params$fontsize <- min(f_size_w, f_size_h)
    ttheme$rowhead$fg_params$fontsize <- min(f_size_w, f_size_h)
  }

  tryCatch(
    expr = {
      gt <- gridExtra::tableGrob(
        d = data,
        theme = ttheme
      ) # ERROR 'data' must be of a vector type, was 'NULL'
      gt$widths <- ((w_unit / sum(w_unit)) * width)
      gt$heights <- rep(grid::unit(as.numeric(height) / 4, grid::unitType(height)), nrow(gt))
      vp <- grid::viewport(
        x = grid::unit(x, "npc") + grid::unit(1, "lines"),
        y = grid::unit(y, "npc") + grid::unit(1.5, "lines"),
        height = height,
        width = width,
        just = c("left", "bottom")
      )
      grid::gList(
        grid::gTree(
          vp = vp,
          children = grid::gList(gt)
        )
      )
    },
    error = function(w) {
      message(paste(
        "Warning: Cox table will not be displayed as there is",
        "not any level to be compared in the arm variable."
      ))
      return(
        grid::gList(
          grid::gTree(
            vp = NULL,
            children = NULL
          )
        )
      )
    }
  )
}

#' Difference test for two proportions
#'
#' @description `r lifecycle::badge("stable")`
#'
#' The analyze function [test_proportion_diff()] creates a layout element to test the difference between two
#' proportions. The primary analysis variable, `vars`, indicates whether a response has occurred for each record. See
#' the `method` parameter for options of methods to use to calculate the p-value. Additionally, a stratification
#' variable can be supplied via the `strata` element of the `variables` argument.
#'
#' @inheritParams argument_convention
#' @param method (`string`)\cr one of `chisq`, `cmh`, `fisher`, or `schouten`; specifies the test used
#'   to calculate the p-value.
#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("test_proportion_diff")`
#'   to see available statistics for this function.
#'
#' @seealso [h_prop_diff_test]
#'
#' @name prop_diff_test
#' @order 1
NULL

#' @describeIn prop_diff_test Statistics function which tests the difference between two proportions.
#'
#' @return
#' * `s_test_proportion_diff()` returns a named `list` with a single item `pval` with an attribute `label`
#'   describing the method used. The p-value tests the null hypothesis that proportions in two groups are the same.
#'
#' @keywords internal
s_test_proportion_diff <- function(df,
                                   .var,
                                   .ref_group,
                                   .in_ref_col,
                                   variables = list(strata = NULL),
                                   method = c("chisq", "schouten", "fisher", "cmh")) {
  method <- match.arg(method)
  y <- list(pval = "")

  if (!.in_ref_col) {
    assert_df_with_variables(df, list(rsp = .var))
    assert_df_with_variables(.ref_group, list(rsp = .var))
    rsp <- factor(
      c(.ref_group[[.var]], df[[.var]]),
      levels = c("TRUE", "FALSE")
    )
    grp <- factor(
      rep(c("ref", "Not-ref"), c(nrow(.ref_group), nrow(df))),
      levels = c("ref", "Not-ref")
    )

    if (!is.null(variables$strata) || method == "cmh") {
      strata <- variables$strata
      checkmate::assert_false(is.null(strata))
      strata_vars <- stats::setNames(as.list(strata), strata)
      assert_df_with_variables(df, strata_vars)
      assert_df_with_variables(.ref_group, strata_vars)
      strata <- c(interaction(.ref_group[strata]), interaction(df[strata]))
    }

    tbl <- switch(method,
      cmh = table(grp, rsp, strata),
      table(grp, rsp)
    )

    y$pval <- switch(method,
      chisq = prop_chisq(tbl),
      cmh = prop_cmh(tbl),
      fisher = prop_fisher(tbl),
      schouten = prop_schouten(tbl)
    )
  }

  y$pval <- formatters::with_label(y$pval, d_test_proportion_diff(method))
  y
}

#' Description of the difference test between two proportions
#'
#' @description `r lifecycle::badge("stable")`
#'
#' This is an auxiliary function that describes the analysis in `s_test_proportion_diff`.
#'
#' @inheritParams s_test_proportion_diff
#'
#' @return A `string` describing the test from which the p-value is derived.
#'
#' @export
d_test_proportion_diff <- function(method) {
  checkmate::assert_string(method)
  meth_part <- switch(method,
    "schouten" = "Chi-Squared Test with Schouten Correction",
    "chisq" = "Chi-Squared Test",
    "cmh" = "Cochran-Mantel-Haenszel Test",
    "fisher" = "Fisher's Exact Test",
    stop(paste(method, "does not have a description"))
  )
  paste0("p-value (", meth_part, ")")
}

#' @describeIn prop_diff_test Formatted analysis function which is used as `afun` in `test_proportion_diff()`.
#'
#' @return
#' * `a_test_proportion_diff()` returns the corresponding list with formatted [rtables::CellValue()].
#'
#' @keywords internal
a_test_proportion_diff <- make_afun(
  s_test_proportion_diff,
  .formats = c(pval = "x.xxxx | (<0.0001)"),
  .indent_mods = c(pval = 1L)
)

#' @describeIn prop_diff_test Layout-creating function which can take statistics function arguments
#'   and additional format arguments. This function is a wrapper for [rtables::analyze()].
#'
#' @return
#' * `test_proportion_diff()` returns a layout object suitable for passing to further layouting functions,
#'   or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
#'   the statistics from `s_test_proportion_diff()` to the table layout.
#'
#' @examples
#' dta <- data.frame(
#'   rsp = sample(c(TRUE, FALSE), 100, TRUE),
#'   grp = factor(rep(c("A", "B"), each = 50)),
#'   strata = factor(rep(c("V", "W", "X", "Y", "Z"), each = 20))
#' )
#'
#' # With `rtables` pipelines.
#' l <- basic_table() %>%
#'   split_cols_by(var = "grp", ref_group = "B") %>%
#'   test_proportion_diff(
#'     vars = "rsp",
#'     method = "cmh", variables = list(strata = "strata")
#'   )
#'
#' build_table(l, df = dta)
#'
#' @export
#' @order 2
test_proportion_diff <- function(lyt,
                                 vars,
                                 variables = list(strata = NULL),
                                 method = c("chisq", "schouten", "fisher", "cmh"),
                                 na_str = default_na_str(),
                                 nested = TRUE,
                                 ...,
                                 var_labels = vars,
                                 show_labels = "hidden",
                                 table_names = vars,
                                 .stats = NULL,
                                 .formats = NULL,
                                 .labels = NULL,
                                 .indent_mods = NULL) {
  extra_args <- list(variables = variables, method = method, ...)

  afun <- make_afun(
    a_test_proportion_diff,
    .stats = .stats,
    .formats = .formats,
    .labels = .labels,
    .indent_mods = .indent_mods
  )
  analyze(
    lyt,
    vars,
    afun = afun,
    var_labels = var_labels,
    na_str = na_str,
    nested = nested,
    extra_args = extra_args,
    show_labels = show_labels,
    table_names = table_names
  )
}

#' Helper functions to test proportion differences
#'
#' Helper functions to implement various tests on the difference between two proportions.
#'
#' @param tbl (`matrix`)\cr matrix with two groups in rows and the binary response (`TRUE`/`FALSE`) in columns.
#'
#' @return A p-value.
#'
#' @seealso [prop_diff_test()] for implementation of these helper functions.
#'
#' @name h_prop_diff_test
NULL

#' @describeIn h_prop_diff_test Performs Chi-Squared test. Internally calls [stats::prop.test()].
#'
#' @keywords internal
prop_chisq <- function(tbl) {
  checkmate::assert_integer(c(ncol(tbl), nrow(tbl)), lower = 2, upper = 2)
  tbl <- tbl[, c("TRUE", "FALSE")]
  if (any(colSums(tbl) == 0)) {
    return(1)
  }
  stats::prop.test(tbl, correct = FALSE)$p.value
}

#' @describeIn h_prop_diff_test Performs stratified Cochran-Mantel-Haenszel test. Internally calls
#'   [stats::mantelhaen.test()]. Note that strata with less than two observations are automatically discarded.
#'
#' @param ary (`array`, 3 dimensions)\cr array with two groups in rows, the binary response
#'   (`TRUE`/`FALSE`) in columns, and the strata in the third dimension.
#'
#' @keywords internal
prop_cmh <- function(ary) {
  checkmate::assert_array(ary)
  checkmate::assert_integer(c(ncol(ary), nrow(ary)), lower = 2, upper = 2)
  checkmate::assert_integer(length(dim(ary)), lower = 3, upper = 3)
  strata_sizes <- apply(ary, MARGIN = 3, sum)
  if (any(strata_sizes < 5)) {
    warning("<5 data points in some strata. CMH test may be incorrect.")
    ary <- ary[, , strata_sizes > 1]
  }

  stats::mantelhaen.test(ary, correct = FALSE)$p.value
}

#' @describeIn h_prop_diff_test Performs the Chi-Squared test with Schouten correction.
#'
#' @seealso Schouten correction is based upon \insertCite{Schouten1980-kd;textual}{tern}.
#'
#' @keywords internal
prop_schouten <- function(tbl) {
  checkmate::assert_integer(c(ncol(tbl), nrow(tbl)), lower = 2, upper = 2)
  tbl <- tbl[, c("TRUE", "FALSE")]
  if (any(colSums(tbl) == 0)) {
    return(1)
  }

  n <- sum(tbl)
  n1 <- sum(tbl[1, ])
  n2 <- sum(tbl[2, ])

  ad <- diag(tbl)
  bc <- diag(apply(tbl, 2, rev))
  ac <- tbl[, 1]
  bd <- tbl[, 2]

  t_schouten <- (n - 1) *
    (abs(prod(ad) - prod(bc)) - 0.5 * min(n1, n2))^2 /
    (n1 * n2 * sum(ac) * sum(bd))

  1 - stats::pchisq(t_schouten, df = 1)
}

#' @describeIn h_prop_diff_test Performs the Fisher's exact test. Internally calls [stats::fisher.test()].
#'
#' @keywords internal
prop_fisher <- function(tbl) {
  checkmate::assert_integer(c(ncol(tbl), nrow(tbl)), lower = 2, upper = 2)
  tbl <- tbl[, c("TRUE", "FALSE")]
  stats::fisher.test(tbl)$p.value
}

#' Count occurrences by grade
#'
#' @description `r lifecycle::badge("stable")`
#'
#' The analyze function [count_occurrences_by_grade()] creates a layout element to calculate occurrence counts by grade.
#'
#' This function analyzes primary analysis variable `var` which indicates toxicity grades. The `id` variable
#' is used to indicate unique subject identifiers (defaults to `USUBJID`). The user can also supply a list of
#' custom groups of grades to analyze via the `grade_groups` parameter. The `remove_single`  argument will
#' remove single grades from the analysis so that *only* grade groups are analyzed.
#'
#' If there are multiple grades recorded for one patient only the highest grade level is counted.
#'
#' The summarize function [summarize_occurrences_by_grade()] performs the same function as
#' [count_occurrences_by_grade()] except it creates content rows, not data rows, to summarize the current table
#' row/column context and operates on the level of the latest row split or the root of the table if no row splits have
#' occurred.
#'
#' @inheritParams argument_convention
#' @param grade_groups (named `list` of `character`)\cr list containing groupings of grades.
#' @param remove_single (`flag`)\cr `TRUE` to not include the elements of one-element grade groups
#'   in the the output list; in this case only the grade groups names will be included in the output. If
#'   `only_grade_groups` is set to `TRUE` this argument is ignored.
#' @param only_grade_groups (`flag`)\cr whether only the specified grade groups should be
#'   included, with individual grade rows removed (`TRUE`), or all grades and grade groups
#'   should be displayed (`FALSE`).
#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("count_occurrences_by_grade")`
#'   to see available statistics for this function.
#'
#' @seealso Relevant helper function [h_append_grade_groups()].
#'
#' @name count_occurrences_by_grade
#' @order 1
NULL

#' Helper function for `s_count_occurrences_by_grade()`
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Helper function for [s_count_occurrences_by_grade()] to insert grade groupings into list with
#' individual grade frequencies. The order of the final result follows the order of `grade_groups`.
#' The elements under any-grade group (if any), i.e. the grade group equal to `refs` will be moved to
#' the end. Grade groups names must be unique.
#'
#' @inheritParams count_occurrences_by_grade
#' @param refs (named `list` of `numeric`)\cr named list where each name corresponds to a reference grade level
#'   and each entry represents a count.
#'
#' @return Formatted list of grade groupings.
#'
#' @examples
#' h_append_grade_groups(
#'   list(
#'     "Any Grade" = as.character(1:5),
#'     "Grade 1-2" = c("1", "2"),
#'     "Grade 3-4" = c("3", "4")
#'   ),
#'   list("1" = 10, "2" = 20, "3" = 30, "4" = 40, "5" = 50)
#' )
#'
#' h_append_grade_groups(
#'   list(
#'     "Any Grade" = as.character(5:1),
#'     "Grade A" = "5",
#'     "Grade B" = c("4", "3")
#'   ),
#'   list("1" = 10, "2" = 20, "3" = 30, "4" = 40, "5" = 50)
#' )
#'
#' h_append_grade_groups(
#'   list(
#'     "Any Grade" = as.character(1:5),
#'     "Grade 1-2" = c("1", "2"),
#'     "Grade 3-4" = c("3", "4")
#'   ),
#'   list("1" = 10, "2" = 5, "3" = 0)
#' )
#'
#' @export
h_append_grade_groups <- function(grade_groups, refs, remove_single = TRUE, only_grade_groups = FALSE) {
  checkmate::assert_list(grade_groups)
  checkmate::assert_list(refs)
  refs_orig <- refs
  elements <- unique(unlist(grade_groups))

  ### compute sums in groups
  grp_sum <- lapply(grade_groups, function(i) do.call(sum, refs[i]))
  if (!checkmate::test_subset(elements, names(refs))) {
    padding_el <- setdiff(elements, names(refs))
    refs[padding_el] <- 0
  }
  result <- c(grp_sum, refs)

  ### order result while keeping grade_groups's ordering
  ordr <- grade_groups

  # elements of any-grade group (if any) will be moved to the end
  is_any <- sapply(grade_groups, setequal, y = names(refs))
  ordr[is_any] <- list(character(0)) # hide elements under any-grade group

  # groups-elements combined sequence
  ordr <- c(lapply(names(ordr), function(g) c(g, ordr[[g]])), recursive = TRUE, use.names = FALSE)
  ordr <- ordr[!duplicated(ordr)]

  # append remaining elements (if any)
  ordr <- union(ordr, unlist(grade_groups[is_any])) # from any-grade group
  ordr <- union(ordr, names(refs)) # from refs

  # remove elements of single-element groups, if any
  if (only_grade_groups) {
    ordr <- intersect(ordr, names(grade_groups))
  } else if (remove_single) {
    is_single <- sapply(grade_groups, length) == 1L
    ordr <- setdiff(ordr, unlist(grade_groups[is_single]))
  }

  # apply the order
  result <- result[ordr]

  # remove groups without any elements in the original refs
  # note: it's OK if groups have 0 value
  keep_grp <- vapply(grade_groups, function(x, rf) {
    any(x %in% rf)
  }, rf = names(refs_orig), logical(1))

  keep_el <- names(result) %in% names(refs_orig) | names(result) %in% names(keep_grp)[keep_grp]
  result <- result[keep_el]

  result
}

#' @describeIn count_occurrences_by_grade Statistics function which counts the
#'  number of patients by highest grade.
#'
#' @return
#' * `s_count_occurrences_by_grade()` returns a list of counts and fractions with one element per grade level or
#'   grade level grouping.
#'
#' @examples
#' s_count_occurrences_by_grade(
#'   df,
#'   .N_col = 10L,
#'   .var = "AETOXGR",
#'   id = "USUBJID",
#'   grade_groups = list("ANY" = levels(df$AETOXGR))
#' )
#'
#' @export
s_count_occurrences_by_grade <- function(df,
                                         .var,
                                         .N_col, # nolint
                                         id = "USUBJID",
                                         grade_groups = list(),
                                         remove_single = TRUE,
                                         only_grade_groups = FALSE,
                                         labelstr = "") {
  assert_valid_factor(df[[.var]])
  assert_df_with_variables(df, list(grade = .var, id = id))

  if (nrow(df) < 1) {
    grade_levels <- levels(df[[.var]])
    l_count <- as.list(rep(0, length(grade_levels)))
    names(l_count) <- grade_levels
  } else {
    if (isTRUE(is.factor(df[[id]]))) {
      assert_valid_factor(df[[id]], any.missing = FALSE)
    } else {
      checkmate::assert_character(df[[id]], min.chars = 1, any.missing = FALSE)
    }
    checkmate::assert_count(.N_col)

    id <- df[[id]]
    grade <- df[[.var]]

    if (!is.ordered(grade)) {
      grade_lbl <- obj_label(grade)
      lvls <- levels(grade)
      if (sum(grepl("^\\d+$", lvls)) %in% c(0, length(lvls))) {
        lvl_ord <- lvls
      } else {
        lvls[!grepl("^\\d+$", lvls)] <- min(as.numeric(lvls[grepl("^\\d+$", lvls)])) - 1
        lvl_ord <- levels(grade)[order(as.numeric(lvls))]
      }
      grade <- formatters::with_label(factor(grade, levels = lvl_ord, ordered = TRUE), grade_lbl)
    }

    missing_lvl <- grepl("missing", tolower(levels(grade)))
    if (any(missing_lvl)) {
      grade <- factor(
        grade,
        levels = c(levels(grade)[!missing_lvl], levels(grade)[missing_lvl]),
        ordered = is.ordered(grade)
      )
    }
    df_max <- stats::aggregate(grade ~ id, FUN = max, drop = FALSE)
    l_count <- as.list(table(df_max$grade))
  }

  if (length(grade_groups) > 0) {
    l_count <- h_append_grade_groups(grade_groups, l_count, remove_single, only_grade_groups)
  }

  l_count_fraction <- lapply(l_count, function(i, denom) c(i, i / denom), denom = .N_col)

  list(
    count_fraction = l_count_fraction
  )
}

#' @describeIn count_occurrences_by_grade Formatted analysis function which is used as `afun`
#'   in `count_occurrences_by_grade()`.
#'
#' @return
#' * `a_count_occurrences_by_grade()` returns the corresponding list with formatted [rtables::CellValue()].
#'
#' @examples
#' #  We need to ungroup `count_fraction` first so that the `rtables` formatting
#' # function `format_count_fraction()` can be applied correctly.
#' afun <- make_afun(a_count_occurrences_by_grade, .ungroup_stats = "count_fraction")
#' afun(
#'   df,
#'   .N_col = 10L,
#'   .var = "AETOXGR",
#'   id = "USUBJID",
#'   grade_groups = list("ANY" = levels(df$AETOXGR))
#' )
#'
#' @export
a_count_occurrences_by_grade <- make_afun(
  s_count_occurrences_by_grade,
  .formats = c("count_fraction" = format_count_fraction_fixed_dp)
)

#' @describeIn count_occurrences_by_grade Layout-creating function which can take statistics function
#'   arguments and additional format arguments. This function is a wrapper for [rtables::analyze()].
#'
#' @return
#' * `count_occurrences_by_grade()` returns a layout object suitable for passing to further layouting functions,
#'   or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
#'   the statistics from `s_count_occurrences_by_grade()` to the table layout.
#'
#' @examples
#' library(dplyr)
#'
#' df <- data.frame(
#'   USUBJID = as.character(c(1:6, 1)),
#'   ARM = factor(c("A", "A", "A", "B", "B", "B", "A"), levels = c("A", "B")),
#'   AETOXGR = factor(c(1, 2, 3, 4, 1, 2, 3), levels = c(1:5)),
#'   AESEV = factor(
#'     x = c("MILD", "MODERATE", "SEVERE", "MILD", "MILD", "MODERATE", "SEVERE"),
#'     levels = c("MILD", "MODERATE", "SEVERE")
#'   ),
#'   stringsAsFactors = FALSE
#' )
#'
#' df_adsl <- df %>%
#'   select(USUBJID, ARM) %>%
#'   unique()
#'
#' # Layout creating function with custom format.
#' basic_table() %>%
#'   split_cols_by("ARM") %>%
#'   add_colcounts() %>%
#'   count_occurrences_by_grade(
#'     var = "AESEV",
#'     .formats = c("count_fraction" = "xx.xx (xx.xx%)")
#'   ) %>%
#'   build_table(df, alt_counts_df = df_adsl)
#'
#' # Define additional grade groupings.
#' grade_groups <- list(
#'   "-Any-" = c("1", "2", "3", "4", "5"),
#'   "Grade 1-2" = c("1", "2"),
#'   "Grade 3-5" = c("3", "4", "5")
#' )
#'
#' basic_table() %>%
#'   split_cols_by("ARM") %>%
#'   add_colcounts() %>%
#'   count_occurrences_by_grade(
#'     var = "AETOXGR",
#'     grade_groups = grade_groups,
#'     only_grade_groups = TRUE
#'   ) %>%
#'   build_table(df, alt_counts_df = df_adsl)
#'
#' @export
#' @order 2
count_occurrences_by_grade <- function(lyt,
                                       var,
                                       id = "USUBJID",
                                       grade_groups = list(),
                                       remove_single = TRUE,
                                       only_grade_groups = FALSE,
                                       var_labels = var,
                                       show_labels = "default",
                                       riskdiff = FALSE,
                                       na_str = default_na_str(),
                                       nested = TRUE,
                                       ...,
                                       table_names = var,
                                       .stats = NULL,
                                       .formats = NULL,
                                       .indent_mods = NULL,
                                       .labels = NULL) {
  checkmate::assert_flag(riskdiff)
  s_args <- list(
    id = id, grade_groups = grade_groups, remove_single = remove_single, only_grade_groups = only_grade_groups, ...
  )

  afun <- make_afun(
    a_count_occurrences_by_grade,
    .stats = .stats,
    .formats = .formats,
    .indent_mods = .indent_mods,
    .ungroup_stats = "count_fraction"
  )

  extra_args <- if (isFALSE(riskdiff)) {
    s_args
  } else {
    list(
      afun = list("s_count_occurrences_by_grade" = afun),
      .stats = .stats,
      .indent_mods = .indent_mods,
      s_args = s_args
    )
  }

  analyze(
    lyt = lyt,
    vars = var,
    var_labels = var_labels,
    show_labels = show_labels,
    afun = ifelse(isFALSE(riskdiff), afun, afun_riskdiff),
    table_names = table_names,
    na_str = na_str,
    nested = nested,
    extra_args = extra_args
  )
}

#' @describeIn count_occurrences_by_grade Layout-creating function which can take content function arguments
#'   and additional format arguments. This function is a wrapper for [rtables::summarize_row_groups()].
#'
#' @return
#' * `summarize_occurrences_by_grade()` returns a layout object suitable for passing to further layouting functions,
#'   or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted content rows
#'   containing the statistics from `s_count_occurrences_by_grade()` to the table layout.
#'
#' @examples
#' # Layout creating function with custom format.
#' basic_table() %>%
#'   add_colcounts() %>%
#'   split_rows_by("ARM", child_labels = "visible", nested = TRUE) %>%
#'   summarize_occurrences_by_grade(
#'     var = "AESEV",
#'     .formats = c("count_fraction" = "xx.xx (xx.xx%)")
#'   ) %>%
#'   build_table(df, alt_counts_df = df_adsl)
#'
#' basic_table() %>%
#'   add_colcounts() %>%
#'   split_rows_by("ARM", child_labels = "visible", nested = TRUE) %>%
#'   summarize_occurrences_by_grade(
#'     var = "AETOXGR",
#'     grade_groups = grade_groups
#'   ) %>%
#'   build_table(df, alt_counts_df = df_adsl)
#'
#' @export
#' @order 3
summarize_occurrences_by_grade <- function(lyt,
                                           var,
                                           id = "USUBJID",
                                           grade_groups = list(),
                                           remove_single = TRUE,
                                           only_grade_groups = FALSE,
                                           na_str = default_na_str(),
                                           ...,
                                           .stats = NULL,
                                           .formats = NULL,
                                           .indent_mods = NULL,
                                           .labels = NULL) {
  extra_args <- list(
    id = id, grade_groups = grade_groups, remove_single = remove_single, only_grade_groups = only_grade_groups, ...
  )

  cfun <- make_afun(
    a_count_occurrences_by_grade,
    .stats = .stats,
    .formats = .formats,
    .labels = .labels,
    .indent_mods = .indent_mods,
    .ungroup_stats = "count_fraction"
  )

  summarize_row_groups(
    lyt = lyt,
    var = var,
    cfun = cfun,
    na_str = na_str,
    extra_args = extra_args
  )
}

#' Control function for subgroup treatment effect pattern (STEP) calculations
#'
#' @description `r lifecycle::badge("stable")`
#'
#' This is an auxiliary function for controlling arguments for STEP calculations.
#'
#' @param biomarker (`numeric` or `NULL`)\cr optional provision of the numeric biomarker variable, which
#'   could be used to infer `bandwidth`, see below.
#' @param use_percentile (`flag`)\cr if `TRUE`, the running windows are created according to
#'   quantiles rather than actual values, i.e. the bandwidth refers to the percentage of data
#'   covered in each window. Suggest `TRUE` if the biomarker variable is not uniformly
#'   distributed.
#' @param bandwidth (`numeric(1)` or `NULL`)\cr indicating the bandwidth of each window.
#'   Depending on the argument `use_percentile`, it can be either the length of actual-value
#'   windows on the real biomarker scale, or percentage windows.
#'   If `use_percentile = TRUE`, it should be a number between 0 and 1.
#'   If `NULL`, treat the bandwidth to be infinity, which means only one global model will be fitted.
#'   By default, `0.25` is used for percentage windows and one quarter of the range of the `biomarker`
#'   variable for actual-value windows.
#' @param degree (`integer(1)`)\cr the degree of polynomial function of the biomarker as an interaction term
#'   with the treatment arm fitted at each window. If 0 (default), then the biomarker variable
#'   is not included in the model fitted in each biomarker window.
#' @param num_points (`integer(1)`)\cr the number of points at which the hazard ratios are estimated. The
#'   smallest number is 2.
#'
#' @return A list of components with the same names as the arguments, except `biomarker` which is
#'   just used to calculate the `bandwidth` in case that actual biomarker windows are requested.
#'
#' @examples
#' # Provide biomarker values and request actual values to be used,
#' # so that bandwidth is chosen from range.
#' control_step(biomarker = 1:10, use_percentile = FALSE)
#'
#' # Use a global model with quadratic biomarker interaction term.
#' control_step(bandwidth = NULL, degree = 2)
#'
#' # Reduce number of points to be used.
#' control_step(num_points = 10)
#'
#' @export
control_step <- function(biomarker = NULL,
                         use_percentile = TRUE,
                         bandwidth,
                         degree = 0L,
                         num_points = 39L) {
  checkmate::assert_numeric(biomarker, null.ok = TRUE)
  checkmate::assert_flag(use_percentile)
  checkmate::assert_int(num_points, lower = 2)
  checkmate::assert_count(degree)

  if (missing(bandwidth)) {
    # Infer bandwidth
    bandwidth <- if (use_percentile) {
      0.25
    } else if (!is.null(biomarker)) {
      diff(range(biomarker, na.rm = TRUE)) / 4
    } else {
      NULL
    }
  } else {
    # Check bandwidth
    if (!is.null(bandwidth)) {
      if (use_percentile) {
        assert_proportion_value(bandwidth)
      } else {
        checkmate::assert_scalar(bandwidth)
        checkmate::assert_true(bandwidth > 0)
      }
    }
  }
  list(
    use_percentile = use_percentile,
    bandwidth = bandwidth,
    degree = as.integer(degree),
    num_points = as.integer(num_points)
  )
}

#' Count number of patients
#'
#' @description `r lifecycle::badge("stable")`
#'
#' The analyze function [analyze_num_patients()] creates a layout element to count total numbers of unique or
#' non-unique patients. The primary analysis variable `vars` is used to uniquely identify patients.
#'
#' The `count_by` variable can be used to identify non-unique patients such that the number of patients with a unique
#' combination of values in `vars` and `count_by` will be returned instead as the `nonunique` statistic. The `required`
#' variable can be used to specify a variable required to be non-missing for the record to be included in the counts.
#'
#' The summarize function [summarize_num_patients()] performs the same function as [analyze_num_patients()] except it
#' creates content rows, not data rows, to summarize the current table row/column context and operates on the level of
#' the latest row split or the root of the table if no row splits have occurred.
#'
#' @inheritParams argument_convention
#' @param required (`character` or `NULL`)\cr name of a variable that is required to be non-missing.
#' @param count_by (`character` or `NULL`)\cr name of a variable to be combined with `vars` when counting
#'   `nonunique` records.
#' @param unique_count_suffix (`flag`)\cr whether the `"(n)"` suffix should be added to `unique_count` labels.
#'   Defaults to `TRUE`.
#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("summarize_num_patients")`
#'   to see available statistics for this function.
#'
#' @name summarize_num_patients
#' @order 1
NULL

#' @describeIn summarize_num_patients Statistics function which counts the number of
#'   unique patients, the corresponding percentage taken with respect to the
#'   total number of patients, and the number of non-unique patients.
#'
#' @param x (`character` or `factor`)\cr vector of patient IDs.
#'
#' @return
#' * `s_num_patients()` returns a named `list` of 3 statistics:
#'   * `unique`: Vector of counts and percentages.
#'   * `nonunique`: Vector of counts.
#'   * `unique_count`: Counts.
#'
#' @examples
#' # Use the statistics function to count number of unique and nonunique patients.
#' s_num_patients(x = as.character(c(1, 1, 1, 2, 4, NA)), labelstr = "", .N_col = 6L)
#' s_num_patients(
#'   x = as.character(c(1, 1, 1, 2, 4, NA)),
#'   labelstr = "",
#'   .N_col = 6L,
#'   count_by = c(1, 1, 2, 1, 1, 1)
#' )
#'
#' @export
s_num_patients <- function(x, labelstr, .N_col, count_by = NULL, unique_count_suffix = TRUE) { # nolint

  checkmate::assert_string(labelstr)
  checkmate::assert_count(.N_col)
  checkmate::assert_multi_class(x, classes = c("factor", "character"))
  checkmate::assert_flag(unique_count_suffix)

  count1 <- n_available(unique(x))
  count2 <- n_available(x)

  if (!is.null(count_by)) {
    checkmate::assert_vector(count_by, len = length(x))
    count2 <- n_available(unique(interaction(x, count_by)))
  }

  out <- list(
    unique = formatters::with_label(c(count1, ifelse(count1 == 0 && .N_col == 0, 0, count1 / .N_col)), labelstr),
    nonunique = formatters::with_label(count2, labelstr),
    unique_count = formatters::with_label(
      count1, ifelse(unique_count_suffix, paste0(labelstr, if (nzchar(labelstr)) " ", "(n)"), labelstr)
    )
  )

  out
}

#' @describeIn summarize_num_patients Statistics function which counts the number of unique patients
#'   in a column (variable), the corresponding percentage taken with respect to the total number of
#'   patients, and the number of non-unique patients in the column.
#'
#' @return
#' * `s_num_patients_content()` returns the same values as `s_num_patients()`.
#'
#' @examples
#' # Count number of unique and non-unique patients.
#'
#' df <- data.frame(
#'   USUBJID = as.character(c(1, 2, 1, 4, NA)),
#'   EVENT = as.character(c(10, 15, 10, 17, 8))
#' )
#' s_num_patients_content(df, .N_col = 5, .var = "USUBJID")
#'
#' df_by_event <- data.frame(
#'   USUBJID = as.character(c(1, 2, 1, 4, NA)),
#'   EVENT = c(10, 15, 10, 17, 8)
#' )
#' s_num_patients_content(df_by_event, .N_col = 5, .var = "USUBJID", count_by = "EVENT")
#'
#' @export
s_num_patients_content <- function(df,
                                   labelstr = "",
                                   .N_col, # nolint
                                   .var,
                                   required = NULL,
                                   count_by = NULL,
                                   unique_count_suffix = TRUE) {
  checkmate::assert_string(.var)
  checkmate::assert_data_frame(df)
  if (is.null(count_by)) {
    assert_df_with_variables(df, list(id = .var))
  } else {
    assert_df_with_variables(df, list(id = .var, count_by = count_by))
  }
  if (!is.null(required)) {
    checkmate::assert_string(required)
    assert_df_with_variables(df, list(required = required))
    df <- df[!is.na(df[[required]]), , drop = FALSE]
  }

  x <- df[[.var]]
  y <- if (is.null(count_by)) NULL else df[[count_by]]

  s_num_patients(
    x = x,
    labelstr = labelstr,
    .N_col = .N_col,
    count_by = y,
    unique_count_suffix = unique_count_suffix
  )
}

c_num_patients <- make_afun(
  s_num_patients_content,
  .stats = c("unique", "nonunique", "unique_count"),
  .formats = c(unique = format_count_fraction_fixed_dp, nonunique = "xx", unique_count = "xx")
)

#' @describeIn summarize_num_patients Layout-creating function which can take statistics function arguments
#'   and additional format arguments. This function is a wrapper for [rtables::summarize_row_groups()].
#'
#' @return
#' * `summarize_num_patients()` returns a layout object suitable for passing to further layouting functions,
#'   or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
#'   the statistics from `s_num_patients_content()` to the table layout.
#'
#' @examples
#' # summarize_num_patients
#' tbl <- basic_table() %>%
#'   split_cols_by("ARM") %>%
#'   split_rows_by("SEX") %>%
#'   summarize_num_patients("USUBJID", .stats = "unique_count") %>%
#'   build_table(df)
#'
#' tbl
#'
#' @export
#' @order 3
summarize_num_patients <- function(lyt,
                                   var,
                                   required = NULL,
                                   count_by = NULL,
                                   unique_count_suffix = TRUE,
                                   na_str = default_na_str(),
                                   .stats = NULL,
                                   .formats = NULL,
                                   .labels = c(
                                     unique = "Number of patients with at least one event",
                                     nonunique = "Number of events"
                                   ),
                                   .indent_mods = 0L,
                                   riskdiff = FALSE,
                                   ...) {
  checkmate::assert_flag(riskdiff)

  if (is.null(.stats)) .stats <- c("unique", "nonunique", "unique_count")
  if (length(.labels) > length(.stats)) .labels <- .labels[names(.labels) %in% .stats]

  s_args <- list(required = required, count_by = count_by, unique_count_suffix = unique_count_suffix, ...)

  cfun <- make_afun(
    c_num_patients,
    .stats = .stats,
    .formats = .formats,
    .labels = .labels
  )

  extra_args <- if (isFALSE(riskdiff)) {
    s_args
  } else {
    list(
      afun = list("s_num_patients_content" = cfun),
      .stats = .stats,
      .indent_mods = .indent_mods,
      s_args = s_args
    )
  }

  summarize_row_groups(
    lyt = lyt,
    var = var,
    cfun = ifelse(isFALSE(riskdiff), cfun, afun_riskdiff),
    na_str = na_str,
    extra_args = extra_args,
    indent_mod = .indent_mods
  )
}

#' @describeIn summarize_num_patients Layout-creating function which can take statistics function arguments
#'   and additional format arguments. This function is a wrapper for [rtables::analyze()].
#'
#' @return
#' * `analyze_num_patients()` returns a layout object suitable for passing to further layouting functions,
#'   or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
#'   the statistics from `s_num_patients_content()` to the table layout.
#'
#' @details In general, functions that starts with `analyze*` are expected to
#'   work like [rtables::analyze()], while functions that starts with `summarize*`
#'   are based upon [rtables::summarize_row_groups()]. The latter provides a
#'   value for each dividing split in the row and column space, but, being it
#'   bound to the fundamental splits, it is repeated by design in every page
#'   when pagination is involved.
#'
#' @note As opposed to [summarize_num_patients()], this function does not repeat the produced rows.
#'
#' @examples
#' df <- data.frame(
#'   USUBJID = as.character(c(1, 2, 1, 4, NA, 6, 6, 8, 9)),
#'   ARM = c("A", "A", "A", "A", "A", "B", "B", "B", "B"),
#'   AGE = c(10, 15, 10, 17, 8, 11, 11, 19, 17),
#'   SEX = c("M", "M", "M", "F", "F", "F", "M", "F", "M")
#' )
#'
#' # analyze_num_patients
#' tbl <- basic_table() %>%
#'   split_cols_by("ARM") %>%
#'   add_colcounts() %>%
#'   analyze_num_patients("USUBJID", .stats = c("unique")) %>%
#'   build_table(df)
#'
#' tbl
#'
#' @export
#' @order 2
analyze_num_patients <- function(lyt,
                                 vars,
                                 required = NULL,
                                 count_by = NULL,
                                 unique_count_suffix = TRUE,
                                 na_str = default_na_str(),
                                 nested = TRUE,
                                 .stats = NULL,
                                 .formats = NULL,
                                 .labels = c(
                                   unique = "Number of patients with at least one event",
                                   nonunique = "Number of events"
                                 ),
                                 show_labels = c("default", "visible", "hidden"),
                                 .indent_mods = 0L,
                                 riskdiff = FALSE,
                                 ...) {
  checkmate::assert_flag(riskdiff)

  if (is.null(.stats)) .stats <- c("unique", "nonunique", "unique_count")
  if (length(.labels) > length(.stats)) .labels <- .labels[names(.labels) %in% .stats]

  s_args <- list(required = required, count_by = count_by, unique_count_suffix = unique_count_suffix, ...)

  afun <- make_afun(
    c_num_patients,
    .stats = .stats,
    .formats = .formats,
    .labels = .labels
  )

  extra_args <- if (isFALSE(riskdiff)) {
    s_args
  } else {
    list(
      afun = list("s_num_patients_content" = afun),
      .stats = .stats,
      .indent_mods = .indent_mods,
      s_args = s_args
    )
  }

  analyze(
    afun = ifelse(isFALSE(riskdiff), afun, afun_riskdiff),
    lyt = lyt,
    vars = vars,
    na_str = na_str,
    nested = nested,
    extra_args = extra_args,
    show_labels = show_labels,
    indent_mod = .indent_mods
  )
}

#' Re-implemented `range()` default S3 method for numerical objects
#'
#' This function returns `c(NA, NA)` instead of `c(-Inf, Inf)` for zero-length data
#' without any warnings.
#'
#' @param x (`numeric`)\cr a sequence of numbers for which the range is computed.
#' @param na.rm (`flag`)\cr flag indicating if `NA` should be omitted.
#' @param finite (`flag`)\cr flag indicating if non-finite elements should be removed.
#'
#' @return A 2-element vector of class `numeric`.
#'
#' @keywords internal
range_noinf <- function(x, na.rm = FALSE, finite = FALSE) { # nolint

  checkmate::assert_numeric(x)

  if (finite) {
    x <- x[is.finite(x)] # removes NAs too
  } else if (na.rm) {
    x <- x[!is.na(x)]
  }

  if (length(x) == 0) {
    rval <- c(NA, NA)
    mode(rval) <- typeof(x)
  } else {
    rval <- c(min(x, na.rm = FALSE), max(x, na.rm = FALSE))
  }

  return(rval)
}

#' Utility function to create label for confidence interval
#'
#' @description `r lifecycle::badge("stable")`
#'
#' @inheritParams argument_convention
#'
#' @return A `string`.
#'
#' @export
f_conf_level <- function(conf_level) {
  assert_proportion_value(conf_level)
  paste0(conf_level * 100, "% CI")
}

#' Utility function to create label for p-value
#'
#' @description `r lifecycle::badge("stable")`
#'
#' @param test_mean (`numeric(1)`)\cr mean value to test under the null hypothesis.
#'
#' @return A `string`.
#'
#' @export
f_pval <- function(test_mean) {
  checkmate::assert_numeric(test_mean, len = 1)
  paste0("p-value (H0: mean = ", test_mean, ")")
}

#' Utility function to return a named list of covariate names
#'
#' @param covariates (`character`)\cr a vector that can contain single variable names (such as
#'   `"X1"`), and/or interaction terms indicated by `"X1 * X2"`.
#'
#' @return A named `list` of `character` vector.
#'
#' @keywords internal
get_covariates <- function(covariates) {
  checkmate::assert_character(covariates)
  cov_vars <- unique(trimws(unlist(strsplit(covariates, "\\*"))))
  stats::setNames(as.list(cov_vars), cov_vars)
}

#' Replicate entries of a vector if required
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Replicate entries of a vector if required.
#'
#' @inheritParams argument_convention
#' @param n (`integer(1)`)\cr number of entries that are needed.
#'
#' @return `x` if it has the required length already or is `NULL`,
#'   otherwise if it is scalar the replicated version of it with `n` entries.
#'
#' @note This function will fail if `x` is not of length `n` and/or is not a scalar.
#'
#' @export
to_n <- function(x, n) {
  if (is.null(x)) {
    NULL
  } else if (length(x) == 1) {
    rep(x, n)
  } else if (length(x) == n) {
    x
  } else {
    stop("dimension mismatch")
  }
}

#' Check element dimension
#'
#' Checks if the elements in `...` have the same dimension.
#'
#' @param ... (`data.frame` or `vector`)\cr any data frames or vectors.
#' @param omit_null (`flag`)\cr whether `NULL` elements in `...` should be omitted from the check.
#'
#' @return A `logical` value.
#'
#' @keywords internal
check_same_n <- function(..., omit_null = TRUE) {
  dots <- list(...)

  n_list <- Map(
    function(x, name) {
      if (is.null(x)) {
        if (omit_null) {
          NA_integer_
        } else {
          stop("arg", name, "is not supposed to be NULL")
        }
      } else if (is.data.frame(x)) {
        nrow(x)
      } else if (is.atomic(x)) {
        length(x)
      } else {
        stop("data structure for ", name, "is currently not supported")
      }
    },
    dots, names(dots)
  )

  n <- stats::na.omit(unlist(n_list))

  if (length(unique(n)) > 1) {
    sel <- which(n != n[1])
    stop("Dimension mismatch:", paste(names(n)[sel], collapse = ", "), " do not have N=", n[1])
  }

  TRUE
}

#' Utility function to check if a float value is equal to another float value
#'
#' Uses `.Machine$double.eps` as the tolerance for the comparison.
#'
#' @param x (`numeric(1)`)\cr a float number.
#' @param y (`numeric(1)`)\cr a float number.
#'
#' @return `TRUE` if identical, otherwise `FALSE`.
#'
#' @keywords internal
.is_equal_float <- function(x, y) {
  checkmate::assert_number(x)
  checkmate::assert_number(y)

  # Define a tolerance
  tolerance <- .Machine$double.eps

  # Check if x is close enough to y
  abs(x - y) < tolerance
}

#' Make names without dots
#'
#' @param nams (`character`)\cr vector of original names.
#'
#' @return A `character` `vector` of proper names, which does not use dots in contrast to [make.names()].
#'
#' @keywords internal
make_names <- function(nams) {
  orig <- make.names(nams)
  gsub(".", "", x = orig, fixed = TRUE)
}

#' Conversion of months to days
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Conversion of months to days. This is an approximative calculation because it
#' considers each month as having an average of 30.4375 days.
#'
#' @param x (`numeric(1)`)\cr time in months.
#'
#' @return A `numeric` vector with the time in days.
#'
#' @examples
#' x <- c(13.25, 8.15, 1, 2.834)
#' month2day(x)
#'
#' @export
month2day <- function(x) {
  checkmate::assert_numeric(x)
  x * 30.4375
}

#' Conversion of days to months
#'
#' @param x (`numeric(1)`)\cr time in days.
#'
#' @return A `numeric` vector with the time in months.
#'
#' @examples
#' x <- c(403, 248, 30, 86)
#' day2month(x)
#'
#' @export
day2month <- function(x) {
  checkmate::assert_numeric(x)
  x / 30.4375
}

#' Return an empty numeric if all elements are `NA`.
#'
#' @param x (`numeric`)\cr vector.
#'
#' @return An empty `numeric` if all elements of `x` are `NA`, otherwise `x`.
#'
#' @examples
#' x <- c(NA, NA, NA)
#' # Internal function - empty_vector_if_na
#' @keywords internal
empty_vector_if_na <- function(x) {
  if (all(is.na(x))) {
    numeric()
  } else {
    x
  }
}

#' Element-wise combination of two vectors
#'
#' @param x (`vector`)\cr first vector to combine.
#' @param y (`vector`)\cr second vector to combine.
#'
#' @return A `list` where each element combines corresponding elements of `x` and `y`.
#'
#' @examples
#' combine_vectors(1:3, 4:6)
#'
#' @export
combine_vectors <- function(x, y) {
  checkmate::assert_vector(x)
  checkmate::assert_vector(y, len = length(x))

  result <- lapply(as.data.frame(rbind(x, y)), `c`)
  names(result) <- NULL
  result
}

#' Extract elements by name
#'
#' This utility function extracts elements from a vector `x` by `names`.
#' Differences to the standard `[` function are:
#'
#' - If `x` is `NULL`, then still always `NULL` is returned (same as in base function).
#' - If `x` is not `NULL`, then the intersection of its names is made with `names` and those
#'   elements are returned. That is, `names` which don't appear in `x` are not returned as `NA`s.
#'
#' @param x (named `vector`)\cr where to extract named elements from.
#' @param names (`character`)\cr vector of names to extract.
#'
#' @return `NULL` if `x` is `NULL`, otherwise the extracted elements from `x`.
#'
#' @keywords internal
extract_by_name <- function(x, names) {
  if (is.null(x)) {
    return(NULL)
  }
  checkmate::assert_named(x)
  checkmate::assert_character(names)
  which_extract <- intersect(names(x), names)
  if (length(which_extract) > 0) {
    x[which_extract]
  } else {
    NULL
  }
}

#' Labels for adverse event baskets
#'
#' @description `r lifecycle::badge("stable")`
#'
#' @param aesi (`character`)\cr vector with standardized MedDRA query name (e.g. `SMQxxNAM`) or customized query
#'   name (e.g. `CQxxNAM`).
#' @param scope (`character`)\cr vector with scope of query (e.g. `SMQxxSC`).
#'
#' @return A `string` with the standard label for the AE basket.
#'
#' @examples
#' adae <- tern_ex_adae
#'
#' # Standardized query label includes scope.
#' aesi_label(adae$SMQ01NAM, scope = adae$SMQ01SC)
#'
#' # Customized query label.
#' aesi_label(adae$CQ01NAM)
#'
#' @export
aesi_label <- function(aesi, scope = NULL) {
  checkmate::assert_character(aesi)
  checkmate::assert_character(scope, null.ok = TRUE)
  aesi_label <- obj_label(aesi)
  aesi <- sas_na(aesi)
  aesi <- unique(aesi)[!is.na(unique(aesi))]

  lbl <- if (length(aesi) == 1 && !is.null(scope)) {
    scope <- sas_na(scope)
    scope <- unique(scope)[!is.na(unique(scope))]
    checkmate::assert_string(scope)
    paste0(aesi, " (", scope, ")")
  } else if (length(aesi) == 1 && is.null(scope)) {
    aesi
  } else {
    aesi_label
  }

  lbl
}

#' Indicate study arm variable in formula
#'
#' We use `study_arm` to indicate the study arm variable in `tern` formulas.
#'
#' @param x arm information
#'
#' @return `x`
#'
#' @keywords internal
study_arm <- function(x) {
  structure(x, varname = deparse(substitute(x)))
}

#' Smooth function with optional grouping
#'
#' @description `r lifecycle::badge("stable")`
#'
#' This produces `loess` smoothed estimates of `y` with Student confidence intervals.
#'
#' @param df (`data.frame`)\cr data set containing all analysis variables.
#' @param x (`string`)\cr x column name.
#' @param y (`string`)\cr y column name.
#' @param groups (`character` or `NULL`)\cr vector with optional grouping variables names.
#' @param level (`proportion`)\cr level of confidence interval to use (0.95 by default).
#'
#' @return A `data.frame` with original `x`, smoothed `y`, `ylow`, and `yhigh`, and
#'   optional `groups` variables formatted as `factor` type.
#'
#' @export
get_smooths <- function(df, x, y, groups = NULL, level = 0.95) {
  checkmate::assert_data_frame(df)
  df_cols <- colnames(df)
  checkmate::assert_string(x)
  checkmate::assert_subset(x, df_cols)
  checkmate::assert_numeric(df[[x]])
  checkmate::assert_string(y)
  checkmate::assert_subset(y, df_cols)
  checkmate::assert_numeric(df[[y]])

  if (!is.null(groups)) {
    checkmate::assert_character(groups)
    checkmate::assert_subset(groups, df_cols)
  }

  smooths <- function(x, y) {
    stats::predict(stats::loess(y ~ x), se = TRUE)
  }

  if (!is.null(groups)) {
    cc <- stats::complete.cases(df[c(x, y, groups)])
    df_c <- df[cc, c(x, y, groups)]
    df_c_ordered <- df_c[do.call("order", as.list(df_c[, groups, drop = FALSE])), , drop = FALSE]
    df_c_g <- data.frame(Map(as.factor, df_c_ordered[groups]))

    df_smooth_raw <-
      by(df_c_ordered, df_c_g, function(d) {
        plx <- smooths(d[[x]], d[[y]])
        data.frame(
          x = d[[x]],
          y = plx$fit,
          ylow = plx$fit - stats::qt(level, plx$df) * plx$se.fit,
          yhigh = plx$fit + stats::qt(level, plx$df) * plx$se.fit
        )
      })

    df_smooth <- do.call(rbind, df_smooth_raw)
    df_smooth[groups] <- df_c_g

    df_smooth
  } else {
    cc <- stats::complete.cases(df[c(x, y)])
    df_c <- df[cc, ]
    plx <- smooths(df_c[[x]], df_c[[y]])

    df_smooth <- data.frame(
      x = df_c[[x]],
      y = plx$fit,
      ylow = plx$fit - stats::qt(level, plx$df) * plx$se.fit,
      yhigh = plx$fit + stats::qt(level, plx$df) * plx$se.fit
    )

    df_smooth
  }
}

#' Number of available (non-missing entries) in a vector
#'
#' Small utility function for better readability.
#'
#' @param x (`vector`)\cr vector in which to count non-missing values.
#'
#' @return Number of non-missing values.
#'
#' @keywords internal
n_available <- function(x) {
  sum(!is.na(x))
}

#' Reapply variable labels
#'
#' This is a helper function that is used in tests.
#'
#' @param x (`vector`)\cr vector of elements that needs new labels.
#' @param varlabels (`character`)\cr vector of labels for `x`.
#' @param ... further parameters to be added to the list.
#'
#' @return `x` with variable labels reapplied.
#'
#' @export
reapply_varlabels <- function(x, varlabels, ...) {
  named_labels <- c(as.list(varlabels), list(...))
  formatters::var_labels(x)[names(named_labels)] <- as.character(named_labels)
  x
}

# Wrapper function of survival::clogit so that when model fitting failed, a more useful message would show
clogit_with_tryCatch <- function(formula, data, ...) { # nolint
  tryCatch(
    survival::clogit(formula = formula, data = data, ...),
    error = function(e) stop("model not built successfully with survival::clogit")
  )
}

#' Confidence interval for mean
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Convenient function for calculating the mean confidence interval. It calculates the arithmetic as well as the
#' geometric mean. It can be used as a `ggplot` helper function for plotting.
#'
#' @inheritParams argument_convention
#' @param n_min (`numeric(1)`)\cr a minimum number of non-missing `x` to estimate the confidence interval for mean.
#' @param gg_helper (`flag`)\cr whether output should be aligned for use with `ggplot`s.
#' @param geom_mean (`flag`)\cr whether the geometric mean should be calculated.
#'
#' @return A named `vector` of values `mean_ci_lwr` and `mean_ci_upr`.
#'
#' @examples
#' stat_mean_ci(sample(10), gg_helper = FALSE)
#'
#' p <- ggplot2::ggplot(mtcars, ggplot2::aes(cyl, mpg)) +
#'   ggplot2::geom_point()
#'
#' p + ggplot2::stat_summary(
#'   fun.data = stat_mean_ci,
#'   geom = "errorbar"
#' )
#'
#' p + ggplot2::stat_summary(
#'   fun.data = stat_mean_ci,
#'   fun.args = list(conf_level = 0.5),
#'   geom = "errorbar"
#' )
#'
#' p + ggplot2::stat_summary(
#'   fun.data = stat_mean_ci,
#'   fun.args = list(conf_level = 0.5, geom_mean = TRUE),
#'   geom = "errorbar"
#' )
#'
#' @export
stat_mean_ci <- function(x,
                         conf_level = 0.95,
                         na.rm = TRUE, # nolint
                         n_min = 2,
                         gg_helper = TRUE,
                         geom_mean = FALSE) {
  if (na.rm) {
    x <- stats::na.omit(x)
  }
  n <- length(x)

  if (!geom_mean) {
    m <- mean(x)
  } else {
    negative_values_exist <- any(is.na(x[!is.na(x)]) <- x[!is.na(x)] <= 0)
    if (negative_values_exist) {
      m <- NA_real_
    } else {
      x <- log(x)
      m <- mean(x)
    }
  }

  if (n < n_min || is.na(m)) {
    ci <- c(mean_ci_lwr = NA_real_, mean_ci_upr = NA_real_)
  } else {
    hci <- stats::qt((1 + conf_level) / 2, df = n - 1) * stats::sd(x) / sqrt(n)
    ci <- c(mean_ci_lwr = m - hci, mean_ci_upr = m + hci)
    if (geom_mean) {
      ci <- exp(ci)
    }
  }

  if (gg_helper) {
    m <- ifelse(is.na(m), NA_real_, m)
    ci <- data.frame(y = ifelse(geom_mean, exp(m), m), ymin = ci[[1]], ymax = ci[[2]])
  }

  return(ci)
}

#' Confidence interval for median
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Convenient function for calculating the median confidence interval. It can be used as a `ggplot` helper
#' function for plotting.
#'
#' @inheritParams argument_convention
#' @param gg_helper (`flag`)\cr whether output should be aligned for use with `ggplot`s.
#'
#' @details This function was adapted from `DescTools/versions/0.99.35/source`
#'
#' @return A named `vector` of values `median_ci_lwr` and `median_ci_upr`.
#'
#' @examples
#' stat_median_ci(sample(10), gg_helper = FALSE)
#'
#' p <- ggplot2::ggplot(mtcars, ggplot2::aes(cyl, mpg)) +
#'   ggplot2::geom_point()
#' p + ggplot2::stat_summary(
#'   fun.data = stat_median_ci,
#'   geom = "errorbar"
#' )
#'
#' @export
stat_median_ci <- function(x,
                           conf_level = 0.95,
                           na.rm = TRUE, # nolint
                           gg_helper = TRUE) {
  x <- unname(x)
  if (na.rm) {
    x <- x[!is.na(x)]
  }
  n <- length(x)
  med <- stats::median(x)

  k <- stats::qbinom(p = (1 - conf_level) / 2, size = n, prob = 0.5, lower.tail = TRUE)

  # k == 0 - for small samples (e.g. n <= 5) ci can be outside the observed range
  if (k == 0 || is.na(med)) {
    ci <- c(median_ci_lwr = NA_real_, median_ci_upr = NA_real_)
    empir_conf_level <- NA_real_
  } else {
    x_sort <- sort(x)
    ci <- c(median_ci_lwr = x_sort[k], median_ci_upr = x_sort[n - k + 1])
    empir_conf_level <- 1 - 2 * stats::pbinom(k - 1, size = n, prob = 0.5)
  }

  if (gg_helper) {
    ci <- data.frame(y = med, ymin = ci[[1]], ymax = ci[[2]])
  }

  attr(ci, "conf_level") <- empir_conf_level

  return(ci)
}

#' p-Value of the mean
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Convenient function for calculating the two-sided p-value of the mean.
#'
#' @inheritParams argument_convention
#' @param n_min (`numeric(1)`)\cr a minimum number of non-missing `x` to estimate the p-value of the mean.
#' @param test_mean (`numeric(1)`)\cr mean value to test under the null hypothesis.
#'
#' @return A p-value.
#'
#' @examples
#' stat_mean_pval(sample(10))
#'
#' stat_mean_pval(rnorm(10), test_mean = 0.5)
#'
#' @export
stat_mean_pval <- function(x,
                           na.rm = TRUE, # nolint
                           n_min = 2,
                           test_mean = 0) {
  if (na.rm) {
    x <- stats::na.omit(x)
  }
  n <- length(x)

  x_mean <- mean(x)
  x_sd <- stats::sd(x)

  if (n < n_min) {
    pv <- c(p_value = NA_real_)
  } else {
    x_se <- stats::sd(x) / sqrt(n)
    ttest <- (x_mean - test_mean) / x_se
    pv <- c(p_value = 2 * stats::pt(-abs(ttest), df = n - 1))
  }

  return(pv)
}

#' Proportion difference and confidence interval
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Function for calculating the proportion (or risk) difference and confidence interval between arm
#' X (reference group) and arm Y. Risk difference is calculated by subtracting cumulative incidence
#' in arm Y from cumulative incidence in arm X.
#'
#' @inheritParams argument_convention
#' @param x (`list` of `integer`)\cr list of number of occurrences in arm X (reference group).
#' @param y (`list` of `integer`)\cr list of number of occurrences in arm Y. Must be of equal length to `x`.
#' @param N_x (`numeric(1)`)\cr total number of records in arm X.
#' @param N_y (`numeric(1)`)\cr total number of records in arm Y.
#' @param list_names (`character`)\cr names of each variable/level corresponding to pair of proportions in
#'   `x` and `y`. Must be of equal length to `x` and `y`.
#' @param pct (`flag`)\cr whether output should be returned as percentages. Defaults to `TRUE`.
#'
#' @return List of proportion differences and CIs corresponding to each pair of number of occurrences in `x` and
#'   `y`. Each list element consists of 3 statistics: proportion difference, CI lower bound, and CI upper bound.
#'
#' @seealso Split function [add_riskdiff()] which, when used as `split_fun` within [rtables::split_cols_by()]
#'   with `riskdiff` argument is set to `TRUE` in subsequent analyze functions, adds a column containing
#'   proportion (risk) difference to an `rtables` layout.
#'
#' @examples
#' stat_propdiff_ci(
#'   x = list(0.375), y = list(0.01), N_x = 5, N_y = 5, list_names = "x", conf_level = 0.9
#' )
#'
#' stat_propdiff_ci(
#'   x = list(0.5, 0.75, 1), y = list(0.25, 0.05, 0.5), N_x = 10, N_y = 20, pct = FALSE
#' )
#'
#' @export
stat_propdiff_ci <- function(x,
                             y,
                             N_x, # nolint
                             N_y, # nolint
                             list_names = NULL,
                             conf_level = 0.95,
                             pct = TRUE) {
  checkmate::assert_list(x, types = "numeric")
  checkmate::assert_list(y, types = "numeric", len = length(x))
  checkmate::assert_character(list_names, len = length(x), null.ok = TRUE)
  rd_list <- lapply(seq_along(x), function(i) {
    p_x <- x[[i]] / N_x
    p_y <- y[[i]] / N_y
    rd_ci <- p_x - p_y + c(-1, 1) * stats::qnorm((1 + conf_level) / 2) *
      sqrt(p_x * (1 - p_x) / N_x + p_y * (1 - p_y) / N_y)
    c(p_x - p_y, rd_ci) * ifelse(pct, 100, 1)
  })
  names(rd_list) <- list_names
  rd_list
}

#' Cumulative counts of numeric variable by thresholds
#'
#' @description `r lifecycle::badge("stable")`
#'
#' The analyze function [count_cumulative()] creates a layout element to calculate cumulative counts of values in a
#' numeric variable that are less than, less or equal to, greater than, or greater or equal to user-specified
#' threshold values.
#'
#' This function analyzes numeric variable `vars` against the threshold values supplied to the `thresholds`
#' argument as a numeric vector. Whether counts should include the threshold values, and whether to count
#' values lower or higher than the threshold values can be set via the `include_eq` and `lower_tail`
#' parameters, respectively.
#'
#' @inheritParams h_count_cumulative
#' @inheritParams argument_convention
#' @param thresholds (`numeric`)\cr vector of cutoff values for the counts.
#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("count_cumulative")`
#'   to see available statistics for this function.
#'
#' @seealso Relevant helper function [h_count_cumulative()], and descriptive function [d_count_cumulative()].
#'
#' @name count_cumulative
#' @order 1
NULL

#' Helper function for `s_count_cumulative()`
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Helper function to calculate count and fraction of `x` values in the lower or upper tail given a threshold.
#'
#' @inheritParams argument_convention
#' @param threshold (`numeric(1)`)\cr a cutoff value as threshold to count values of `x`.
#' @param lower_tail (`flag`)\cr whether to count lower tail, default is `TRUE`.
#' @param include_eq (`flag`)\cr whether to include value equal to the `threshold` in
#'   count, default is `TRUE`.
#'
#' @return A named vector with items:
#'   * `count`: the count of values less than, less or equal to, greater than, or greater or equal to a threshold
#'     of user specification.
#'   * `fraction`: the fraction of the count.
#'
#' @seealso [count_cumulative]
#'
#' @examples
#' set.seed(1, kind = "Mersenne-Twister")
#' x <- c(sample(1:10, 10), NA)
#' .N_col <- length(x)
#'
#' h_count_cumulative(x, 5, .N_col = .N_col)
#' h_count_cumulative(x, 5, lower_tail = FALSE, include_eq = FALSE, na.rm = FALSE, .N_col = .N_col)
#' h_count_cumulative(x, 0, lower_tail = FALSE, .N_col = .N_col)
#' h_count_cumulative(x, 100, lower_tail = FALSE, .N_col = .N_col)
#'
#' @export
h_count_cumulative <- function(x,
                               threshold,
                               lower_tail = TRUE,
                               include_eq = TRUE,
                               na.rm = TRUE, # nolint
                               .N_col) { # nolint
  checkmate::assert_numeric(x)
  checkmate::assert_numeric(threshold)
  checkmate::assert_numeric(.N_col)
  checkmate::assert_flag(lower_tail)
  checkmate::assert_flag(include_eq)
  checkmate::assert_flag(na.rm)

  is_keep <- if (na.rm) !is.na(x) else rep(TRUE, length(x))
  count <- if (lower_tail && include_eq) {
    length(x[is_keep & x <= threshold])
  } else if (lower_tail && !include_eq) {
    length(x[is_keep & x < threshold])
  } else if (!lower_tail && include_eq) {
    length(x[is_keep & x >= threshold])
  } else if (!lower_tail && !include_eq) {
    length(x[is_keep & x > threshold])
  }

  result <- c(count = count, fraction = count / .N_col)
  result
}

#' Description of cumulative count
#'
#' @description `r lifecycle::badge("stable")`
#'
#' This is a helper function that describes the analysis in [s_count_cumulative()].
#'
#' @inheritParams h_count_cumulative
#'
#' @return Labels for [s_count_cumulative()].
#'
#' @export
d_count_cumulative <- function(threshold, lower_tail = TRUE, include_eq = TRUE) {
  checkmate::assert_numeric(threshold)
  lg <- if (lower_tail) "<" else ">"
  eq <- if (include_eq) "=" else ""
  paste0(lg, eq, " ", threshold)
}

#' @describeIn count_cumulative Statistics function that produces a named list given a numeric vector of thresholds.
#'
#' @return
#' * `s_count_cumulative()` returns a named list of `count_fraction`s: a list with each `thresholds` value as a
#'   component, each component containing a vector for the count and fraction.
#'
#' @keywords internal
s_count_cumulative <- function(x,
                               thresholds,
                               lower_tail = TRUE,
                               include_eq = TRUE,
                               .N_col, # nolint
                               ...) {
  checkmate::assert_numeric(thresholds, min.len = 1, any.missing = FALSE)

  count_fraction_list <- Map(function(thres) {
    result <- h_count_cumulative(x, thres, lower_tail, include_eq, .N_col = .N_col, ...)
    label <- d_count_cumulative(thres, lower_tail, include_eq)
    formatters::with_label(result, label)
  }, thresholds)

  names(count_fraction_list) <- thresholds
  list(count_fraction = count_fraction_list)
}

#' @describeIn count_cumulative Formatted analysis function which is used as `afun`
#'   in `count_cumulative()`.
#'
#' @return
#' * `a_count_cumulative()` returns the corresponding list with formatted [rtables::CellValue()].
#'
#' @keywords internal
a_count_cumulative <- make_afun(
  s_count_cumulative,
  .formats = c(count_fraction = format_count_fraction)
)

#' @describeIn count_cumulative Layout-creating function which can take statistics function arguments
#'   and additional format arguments. This function is a wrapper for [rtables::analyze()].
#'
#' @return
#' * `count_cumulative()` returns a layout object suitable for passing to further layouting functions,
#'   or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
#'   the statistics from `s_count_cumulative()` to the table layout.
#'
#' @examples
#' basic_table() %>%
#'   split_cols_by("ARM") %>%
#'   add_colcounts() %>%
#'   count_cumulative(
#'     vars = "AGE",
#'     thresholds = c(40, 60)
#'   ) %>%
#'   build_table(tern_ex_adsl)
#'
#' @export
#' @order 2
count_cumulative <- function(lyt,
                             vars,
                             thresholds,
                             lower_tail = TRUE,
                             include_eq = TRUE,
                             var_labels = vars,
                             show_labels = "visible",
                             na_str = default_na_str(),
                             nested = TRUE,
                             ...,
                             table_names = vars,
                             .stats = NULL,
                             .formats = NULL,
                             .labels = NULL,
                             .indent_mods = NULL) {
  extra_args <- list(thresholds = thresholds, lower_tail = lower_tail, include_eq = include_eq, ...)

  afun <- make_afun(
    a_count_cumulative,
    .stats = .stats,
    .formats = .formats,
    .labels = .labels,
    .indent_mods = .indent_mods,
    .ungroup_stats = "count_fraction"
  )
  analyze(
    lyt,
    vars,
    afun = afun,
    na_str = na_str,
    table_names = table_names,
    var_labels = var_labels,
    show_labels = show_labels,
    nested = nested,
    extra_args = extra_args
  )
}

#' Convert `rtable` objects to `ggplot` objects
#'
#' @description `r lifecycle::badge("experimental")`
#'
#' Given a [rtables::rtable()] object, performs basic conversion to a [ggplot2::ggplot()] object built using
#' functions from the `ggplot2` package. Any table titles and/or footnotes are ignored.
#'
#' @param tbl (`VTableTree`)\cr `rtables` table object.
#' @param fontsize (`numeric(1)`)\cr font size.
#' @param colwidths (`numeric` or `NULL`)\cr a vector of column widths. Each element's position in
#'   `colwidths` corresponds to the column of `tbl` in the same position. If `NULL`, column widths
#'   are calculated according to maximum number of characters per column.
#' @param lbl_col_padding (`numeric`)\cr additional padding to use when calculating spacing between
#'   the first (label) column and the second column of `tbl`. If `colwidths` is specified,
#'   the width of the first column becomes `colwidths[1] + lbl_col_padding`. Defaults to 0.
#'
#' @return A `ggplot` object.
#'
#' @examples
#' dta <- data.frame(
#'   ARM     = rep(LETTERS[1:3], rep(6, 3)),
#'   AVISIT  = rep(paste0("V", 1:3), 6),
#'   AVAL    = c(9:1, rep(NA, 9))
#' )
#'
#' lyt <- basic_table() %>%
#'   split_cols_by(var = "ARM") %>%
#'   split_rows_by(var = "AVISIT") %>%
#'   analyze_vars(vars = "AVAL")
#'
#' tbl <- build_table(lyt, df = dta)
#'
#' rtable2gg(tbl)
#'
#' rtable2gg(tbl, fontsize = 15, colwidths = c(2, 1, 1, 1))
#'
#' @export
rtable2gg <- function(tbl, fontsize = 12, colwidths = NULL, lbl_col_padding = 0) {
  mat <- rtables::matrix_form(tbl, indent_rownames = TRUE)
  mat_strings <- formatters::mf_strings(mat)
  mat_aligns <- formatters::mf_aligns(mat)
  mat_indent <- formatters::mf_rinfo(mat)$indent
  mat_display <- formatters::mf_display(mat)
  nlines_hdr <- formatters::mf_nlheader(mat)
  shared_hdr_rows <- which(apply(mat_display, 1, function(x) (any(!x))))

  tbl_df <- data.frame(mat_strings)
  body_rows <- seq(nlines_hdr + 1, nrow(tbl_df))
  mat_aligns <- apply(mat_aligns, 1:2, function(x) if (x == "left") 0 else if (x == "right") 1 else 0.5)

  # Apply indentation in first column
  tbl_df[body_rows, 1] <- sapply(body_rows, function(i) {
    ind_i <- mat_indent[i - nlines_hdr] * 4
    if (ind_i > 0) paste0(paste(rep(" ", ind_i), collapse = ""), tbl_df[i, 1]) else tbl_df[i, 1]
  })

  # Get column widths
  if (is.null(colwidths)) {
    colwidths <- apply(tbl_df, 2, function(x) max(nchar(x))) + 1
  }
  tot_width <- sum(colwidths) + lbl_col_padding

  if (length(shared_hdr_rows) > 0) {
    tbl_df <- tbl_df[-shared_hdr_rows, ]
    mat_aligns <- mat_aligns[-shared_hdr_rows, ]
  }

  res <- ggplot(data = tbl_df) +
    theme_void() +
    scale_x_continuous(limits = c(0, tot_width)) +
    scale_y_continuous(limits = c(0, nrow(mat_strings))) +
    annotate(
      "segment",
      x = 0, xend = tot_width,
      y = nrow(mat_strings) - nlines_hdr + 0.5, yend = nrow(mat_strings) - nlines_hdr + 0.5
    )

  # If header content spans multiple columns, center over these columns
  if (length(shared_hdr_rows) > 0) {
    mat_strings[shared_hdr_rows, ] <- trimws(mat_strings[shared_hdr_rows, ])
    for (hr in shared_hdr_rows) {
      hdr_lbls <- mat_strings[1:hr, mat_display[hr, -1]]
      hdr_lbls <- matrix(hdr_lbls[nzchar(hdr_lbls)], nrow = hr)
      for (idx_hl in seq_len(ncol(hdr_lbls))) {
        cur_lbl <- tail(hdr_lbls[, idx_hl], 1)
        which_cols <- if (hr == 1) {
          which(mat_strings[hr, ] == hdr_lbls[idx_hl])
        } else { # for >2 col splits, only print labels for each unique combo of nested columns
          which(
            apply(mat_strings[1:hr, ], 2, function(x) all(x == hdr_lbls[1:hr, idx_hl]))
          )
        }
        line_pos <- c(
          sum(colwidths[1:(which_cols[1] - 1)]) + 1 + lbl_col_padding,
          sum(colwidths[1:max(which_cols)]) - 1 + lbl_col_padding
        )

        res <- res +
          annotate(
            "text",
            x = mean(line_pos),
            y = nrow(mat_strings) + 1 - hr,
            label = cur_lbl,
            size = fontsize / .pt
          ) +
          annotate(
            "segment",
            x = line_pos[1],
            xend = line_pos[2],
            y = nrow(mat_strings) - hr + 0.5,
            yend = nrow(mat_strings) - hr + 0.5
          )
      }
    }
  }

  # Add table columns
  for (i in seq_len(ncol(tbl_df))) {
    res <- res + annotate(
      "text",
      x = if (i == 1) 0 else sum(colwidths[1:i]) - 0.5 * colwidths[i] + lbl_col_padding,
      y = rev(seq_len(nrow(tbl_df))),
      label = tbl_df[, i],
      hjust = mat_aligns[, i],
      size = fontsize / .pt
    )
  }

  res
}

#' Convert `data.frame` object to `ggplot` object
#'
#' @description `r lifecycle::badge("experimental")`
#'
#' Given a `data.frame` object, performs basic conversion to a [ggplot2::ggplot()] object built using
#' functions from the `ggplot2` package.
#'
#' @param df (`data.frame`)\cr a data frame.
#' @param colwidths (`numeric` or `NULL`)\cr a vector of column widths. Each element's position in
#'   `colwidths` corresponds to the column of `df` in the same position. If `NULL`, column widths
#'   are calculated according to maximum number of characters per column.
#' @param font_size (`numeric(1)`)\cr font size.
#' @param col_labels (`flag`)\cr whether the column names (labels) of `df` should be used as the first row
#'   of the output table.
#' @param col_lab_fontface (`string`)\cr font face to apply to the first row (of column labels
#'   if `col_labels = TRUE`). Defaults to `"bold"`.
#' @param hline (`flag`)\cr whether a horizontal line should be printed below the first row of the table.
#' @param bg_fill (`string`)\cr table background fill color.
#'
#' @return A `ggplot` object.
#'
#' @examples
#' \dontrun{
#' df2gg(head(iris, 5))
#'
#' df2gg(head(iris, 5), font_size = 15, colwidths = c(1, 1, 1, 1, 1))
#' }
#' @keywords internal
df2gg <- function(df,
                  colwidths = NULL,
                  font_size = 10,
                  col_labels = TRUE,
                  col_lab_fontface = "bold",
                  hline = TRUE,
                  bg_fill = NULL) {
  # convert to text
  df <- as.data.frame(apply(df, 1:2, function(x) if (is.na(x)) "NA" else as.character(x)))

  if (col_labels) {
    df <- as.matrix(df)
    df <- rbind(colnames(df), df)
  }

  # Get column widths
  if (is.null(colwidths)) {
    colwidths <- apply(df, 2, function(x) max(nchar(x), na.rm = TRUE))
  }
  tot_width <- sum(colwidths)

  res <- ggplot(data = df) +
    theme_void() +
    scale_x_continuous(limits = c(0, tot_width)) +
    scale_y_continuous(limits = c(1, nrow(df)))

  if (!is.null(bg_fill)) res <- res + theme(plot.background = element_rect(fill = bg_fill))

  if (hline) {
    res <- res +
      annotate(
        "segment",
        x = 0 + 0.2 * colwidths[2], xend = tot_width - 0.1 * tail(colwidths, 1),
        y = nrow(df) - 0.5, yend = nrow(df) - 0.5
      )
  }

  for (i in seq_len(ncol(df))) {
    line_pos <- c(
      if (i == 1) 0 else sum(colwidths[1:(i - 1)]),
      sum(colwidths[1:i])
    )
    res <- res +
      annotate(
        "text",
        x = mean(line_pos),
        y = rev(seq_len(nrow(df))),
        label = df[, i],
        size = font_size / .pt,
        fontface = if (col_labels) {
          c(col_lab_fontface, rep("plain", nrow(df) - 1))
        } else {
          rep("plain", nrow(df))
        }
      )
  }

  res
}

#' Helper functions for multivariate logistic regression
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Helper functions used in calculations for logistic regression.
#'
#' @inheritParams argument_convention
#' @param fit_glm (`glm`)\cr logistic regression model fitted by [stats::glm()] with "binomial" family.
#'   Limited functionality is also available for conditional logistic regression models fitted by
#'   [survival::clogit()], currently this is used only by [extract_rsp_biomarkers()].
#' @param x (`character`)\cr a variable or interaction term in `fit_glm` (depending on the helper function used).
#'
#' @examples
#' library(dplyr)
#' library(broom)
#'
#' adrs_f <- tern_ex_adrs %>%
#'   filter(PARAMCD == "BESRSPI") %>%
#'   filter(RACE %in% c("ASIAN", "WHITE", "BLACK OR AFRICAN AMERICAN")) %>%
#'   mutate(
#'     Response = case_when(AVALC %in% c("PR", "CR") ~ 1, TRUE ~ 0),
#'     RACE = factor(RACE),
#'     SEX = factor(SEX)
#'   )
#' formatters::var_labels(adrs_f) <- c(formatters::var_labels(tern_ex_adrs), Response = "Response")
#' mod1 <- fit_logistic(
#'   data = adrs_f,
#'   variables = list(
#'     response = "Response",
#'     arm = "ARMCD",
#'     covariates = c("AGE", "RACE")
#'   )
#' )
#' mod2 <- fit_logistic(
#'   data = adrs_f,
#'   variables = list(
#'     response = "Response",
#'     arm = "ARMCD",
#'     covariates = c("AGE", "RACE"),
#'     interaction = "AGE"
#'   )
#' )
#'
#' @name h_logistic_regression
NULL

#' @describeIn h_logistic_regression Helper function to extract interaction variable names from a fitted
#'   model assuming only one interaction term.
#'
#' @return Vector of names of interaction variables.
#'
#' @export
h_get_interaction_vars <- function(fit_glm) {
  checkmate::assert_class(fit_glm, "glm")
  terms_name <- attr(stats::terms(fit_glm), "term.labels")
  terms_order <- attr(stats::terms(fit_glm), "order")
  interaction_term <- terms_name[terms_order == 2]
  checkmate::assert_string(interaction_term)
  strsplit(interaction_term, split = ":")[[1]]
}

#' @describeIn h_logistic_regression Helper function to get the right coefficient name from the
#'   interaction variable names and the given levels. The main value here is that the order
#'   of first and second variable is checked in the `interaction_vars` input.
#'
#' @param interaction_vars (`character(2)`)\cr interaction variable names.
#' @param first_var_with_level (`character(2)`)\cr the first variable name with the interaction level.
#' @param second_var_with_level (`character(2)`)\cr the second variable name with the interaction level.
#'
#' @return Name of coefficient.
#'
#' @export
h_interaction_coef_name <- function(interaction_vars,
                                    first_var_with_level,
                                    second_var_with_level) {
  checkmate::assert_character(interaction_vars, len = 2, any.missing = FALSE)
  checkmate::assert_character(first_var_with_level, len = 2, any.missing = FALSE)
  checkmate::assert_character(second_var_with_level, len = 2, any.missing = FALSE)
  checkmate::assert_subset(c(first_var_with_level[1], second_var_with_level[1]), interaction_vars)

  first_name <- paste(first_var_with_level, collapse = "")
  second_name <- paste(second_var_with_level, collapse = "")
  if (first_var_with_level[1] == interaction_vars[1]) {
    paste(first_name, second_name, sep = ":")
  } else if (second_var_with_level[1] == interaction_vars[1]) {
    paste(second_name, first_name, sep = ":")
  }
}

#' @describeIn h_logistic_regression Helper function to calculate the odds ratio estimates
#'   for the case when both the odds ratio and the interaction variable are categorical.
#'
#' @param odds_ratio_var (`string`)\cr the odds ratio variable.
#' @param interaction_var (`string`)\cr the interaction variable.
#'
#' @return Odds ratio.
#'
#' @export
h_or_cat_interaction <- function(odds_ratio_var,
                                 interaction_var,
                                 fit_glm,
                                 conf_level = 0.95) {
  interaction_vars <- h_get_interaction_vars(fit_glm)
  checkmate::assert_string(odds_ratio_var)
  checkmate::assert_string(interaction_var)
  checkmate::assert_subset(c(odds_ratio_var, interaction_var), interaction_vars)
  checkmate::assert_vector(interaction_vars, len = 2)

  xs_level <- fit_glm$xlevels
  xs_coef <- stats::coef(fit_glm)
  xs_vcov <- stats::vcov(fit_glm)
  y <- list()
  for (var_level in xs_level[[odds_ratio_var]][-1]) {
    x <- list()
    for (ref_level in xs_level[[interaction_var]]) {
      coef_names <- paste0(odds_ratio_var, var_level)
      if (ref_level != xs_level[[interaction_var]][1]) {
        interaction_coef_name <- h_interaction_coef_name(
          interaction_vars,
          c(odds_ratio_var, var_level),
          c(interaction_var, ref_level)
        )
        coef_names <- c(
          coef_names,
          interaction_coef_name
        )
      }
      if (length(coef_names) > 1) {
        ones <- t(c(1, 1))
        est <- as.numeric(ones %*% xs_coef[coef_names])
        se <- sqrt(as.numeric(ones %*% xs_vcov[coef_names, coef_names] %*% t(ones)))
      } else {
        est <- xs_coef[coef_names]
        se <- sqrt(as.numeric(xs_vcov[coef_names, coef_names]))
      }
      or <- exp(est)
      ci <- exp(est + c(lcl = -1, ucl = 1) * stats::qnorm((1 + conf_level) / 2) * se)
      x[[ref_level]] <- list(or = or, ci = ci)
    }
    y[[var_level]] <- x
  }
  y
}

#' @describeIn h_logistic_regression Helper function to calculate the odds ratio estimates
#'   for the case when either the odds ratio or the interaction variable is continuous.
#'
#' @param at (`numeric` or `NULL`)\cr optional values for the interaction variable. Otherwise
#'   the median is used.
#'
#' @return Odds ratio.
#'
#' @note We don't provide a function for the case when both variables are continuous because
#'   this does not arise in this table, as the treatment arm variable will always be involved
#'   and categorical.
#'
#' @export
h_or_cont_interaction <- function(odds_ratio_var,
                                  interaction_var,
                                  fit_glm,
                                  at = NULL,
                                  conf_level = 0.95) {
  interaction_vars <- h_get_interaction_vars(fit_glm)
  checkmate::assert_string(odds_ratio_var)
  checkmate::assert_string(interaction_var)
  checkmate::assert_subset(c(odds_ratio_var, interaction_var), interaction_vars)
  checkmate::assert_vector(interaction_vars, len = 2)
  checkmate::assert_numeric(at, min.len = 1, null.ok = TRUE, any.missing = FALSE)
  xs_level <- fit_glm$xlevels
  xs_coef <- stats::coef(fit_glm)
  xs_vcov <- stats::vcov(fit_glm)
  xs_class <- attr(fit_glm$terms, "dataClasses")
  model_data <- fit_glm$model
  if (!is.null(at)) {
    checkmate::assert_set_equal(xs_class[interaction_var], "numeric")
  }
  y <- list()
  if (xs_class[interaction_var] == "numeric") {
    if (is.null(at)) {
      at <- ceiling(stats::median(model_data[[interaction_var]]))
    }

    for (var_level in xs_level[[odds_ratio_var]][-1]) {
      x <- list()
      for (increment in at) {
        coef_names <- paste0(odds_ratio_var, var_level)
        if (increment != 0) {
          interaction_coef_name <- h_interaction_coef_name(
            interaction_vars,
            c(odds_ratio_var, var_level),
            c(interaction_var, "")
          )
          coef_names <- c(
            coef_names,
            interaction_coef_name
          )
        }
        if (length(coef_names) > 1) {
          xvec <- t(c(1, increment))
          est <- as.numeric(xvec %*% xs_coef[coef_names])
          se <- sqrt(as.numeric(xvec %*% xs_vcov[coef_names, coef_names] %*% t(xvec)))
        } else {
          est <- xs_coef[coef_names]
          se <- sqrt(as.numeric(xs_vcov[coef_names, coef_names]))
        }
        or <- exp(est)
        ci <- exp(est + c(lcl = -1, ucl = 1) * stats::qnorm((1 + conf_level) / 2) * se)
        x[[as.character(increment)]] <- list(or = or, ci = ci)
      }
      y[[var_level]] <- x
    }
  } else {
    checkmate::assert_set_equal(xs_class[odds_ratio_var], "numeric")
    checkmate::assert_set_equal(xs_class[interaction_var], "factor")
    for (var_level in xs_level[[interaction_var]]) {
      coef_names <- odds_ratio_var
      if (var_level != xs_level[[interaction_var]][1]) {
        interaction_coef_name <- h_interaction_coef_name(
          interaction_vars,
          c(odds_ratio_var, ""),
          c(interaction_var, var_level)
        )
        coef_names <- c(
          coef_names,
          interaction_coef_name
        )
      }
      if (length(coef_names) > 1) {
        xvec <- t(c(1, 1))
        est <- as.numeric(xvec %*% xs_coef[coef_names])
        se <- sqrt(as.numeric(xvec %*% xs_vcov[coef_names, coef_names] %*% t(xvec)))
      } else {
        est <- xs_coef[coef_names]
        se <- sqrt(as.numeric(xs_vcov[coef_names, coef_names]))
      }
      or <- exp(est)
      ci <- exp(est + c(lcl = -1, ucl = 1) * stats::qnorm((1 + conf_level) / 2) * se)
      y[[var_level]] <- list(or = or, ci = ci)
    }
  }
  y
}

#' @describeIn h_logistic_regression Helper function to calculate the odds ratio estimates
#'   in case of an interaction. This is a wrapper for [h_or_cont_interaction()] and
#'   [h_or_cat_interaction()].
#'
#' @return Odds ratio.
#'
#' @export
h_or_interaction <- function(odds_ratio_var,
                             interaction_var,
                             fit_glm,
                             at = NULL,
                             conf_level = 0.95) {
  xs_class <- attr(fit_glm$terms, "dataClasses")
  if (any(xs_class[c(odds_ratio_var, interaction_var)] == "numeric")) {
    h_or_cont_interaction(
      odds_ratio_var,
      interaction_var,
      fit_glm,
      at = at,
      conf_level = conf_level
    )
  } else if (all(xs_class[c(odds_ratio_var, interaction_var)] == "factor")) {
    h_or_cat_interaction(
      odds_ratio_var,
      interaction_var,
      fit_glm,
      conf_level = conf_level
    )
  } else {
    stop("wrong interaction variable class, the interaction variable is not a numeric nor a factor")
  }
}

#' @describeIn h_logistic_regression Helper function to construct term labels from simple terms and the table
#'   of numbers of patients.
#'
#' @param terms (`character`)\cr simple terms.
#' @param table (`table`)\cr table containing numbers for terms.
#'
#' @return Term labels containing numbers of patients.
#'
#' @export
h_simple_term_labels <- function(terms,
                                 table) {
  checkmate::assert_true(is.table(table))
  checkmate::assert_multi_class(terms, classes = c("factor", "character"))
  terms <- as.character(terms)
  term_n <- table[terms]
  paste0(terms, ", n = ", term_n)
}

#' @describeIn h_logistic_regression Helper function to construct term labels from interaction terms and the table
#'   of numbers of patients.
#'
#' @param terms1 (`character`)\cr terms for first dimension (rows).
#' @param terms2 (`character`)\cr terms for second dimension (rows).
#' @param any (`flag`)\cr whether any of `term1` and `term2` can be fulfilled to count the
#'   number of patients. In that case they can only be scalar (strings).
#'
#' @return Term labels containing numbers of patients.
#'
#' @export
h_interaction_term_labels <- function(terms1,
                                      terms2,
                                      table,
                                      any = FALSE) {
  checkmate::assert_true(is.table(table))
  checkmate::assert_flag(any)
  checkmate::assert_multi_class(terms1, classes = c("factor", "character"))
  checkmate::assert_multi_class(terms2, classes = c("factor", "character"))
  terms1 <- as.character(terms1)
  terms2 <- as.character(terms2)
  if (any) {
    checkmate::assert_scalar(terms1)
    checkmate::assert_scalar(terms2)
    paste0(
      terms1, " or ", terms2, ", n = ",
      # Note that we double count in the initial sum the cell [terms1, terms2], therefore subtract.
      sum(c(table[terms1, ], table[, terms2])) - table[terms1, terms2]
    )
  } else {
    term_n <- table[cbind(terms1, terms2)]
    paste0(terms1, " * ", terms2, ", n = ", term_n)
  }
}

#' @describeIn h_logistic_regression Helper function to tabulate the main effect
#'   results of a (conditional) logistic regression model.
#'
#' @return Tabulated main effect results from a logistic regression model.
#'
#' @examples
#' h_glm_simple_term_extract("AGE", mod1)
#' h_glm_simple_term_extract("ARMCD", mod1)
#'
#' @export
h_glm_simple_term_extract <- function(x, fit_glm) {
  checkmate::assert_multi_class(fit_glm, c("glm", "clogit"))
  checkmate::assert_string(x)

  xs_class <- attr(fit_glm$terms, "dataClasses")
  xs_level <- fit_glm$xlevels
  xs_coef <- summary(fit_glm)$coefficients
  stats <- if (inherits(fit_glm, "glm")) {
    c("estimate" = "Estimate", "std_error" = "Std. Error", "pvalue" = "Pr(>|z|)")
  } else {
    c("estimate" = "coef", "std_error" = "se(coef)", "pvalue" = "Pr(>|z|)")
  }
  # Make sure x is not an interaction term.
  checkmate::assert_subset(x, names(xs_class))
  x_sel <- if (xs_class[x] == "numeric") x else paste0(x, xs_level[[x]][-1])
  x_stats <- as.data.frame(xs_coef[x_sel, stats, drop = FALSE], stringsAsFactors = FALSE)
  colnames(x_stats) <- names(stats)
  x_stats$estimate <- as.list(x_stats$estimate)
  x_stats$std_error <- as.list(x_stats$std_error)
  x_stats$pvalue <- as.list(x_stats$pvalue)
  x_stats$df <- as.list(1)
  if (xs_class[x] == "numeric") {
    x_stats$term <- x
    x_stats$term_label <- if (inherits(fit_glm, "glm")) {
      formatters::var_labels(fit_glm$data[x], fill = TRUE)
    } else {
      # We just fill in here with the `term` itself as we don't have the data available.
      x
    }
    x_stats$is_variable_summary <- FALSE
    x_stats$is_term_summary <- TRUE
  } else {
    checkmate::assert_class(fit_glm, "glm")
    # The reason is that we don't have the original data set in the `clogit` object
    # and therefore cannot determine the `x_numbers` here.
    x_numbers <- table(fit_glm$data[[x]])
    x_stats$term <- xs_level[[x]][-1]
    x_stats$term_label <- h_simple_term_labels(x_stats$term, x_numbers)
    x_stats$is_variable_summary <- FALSE
    x_stats$is_term_summary <- TRUE
    main_effects <- car::Anova(fit_glm, type = 3, test.statistic = "Wald")
    x_main <- data.frame(
      pvalue = main_effects[x, "Pr(>Chisq)", drop = TRUE],
      term = xs_level[[x]][1],
      term_label = paste("Reference", h_simple_term_labels(xs_level[[x]][1], x_numbers)),
      df = main_effects[x, "Df", drop = TRUE],
      stringsAsFactors = FALSE
    )
    x_main$pvalue <- as.list(x_main$pvalue)
    x_main$df <- as.list(x_main$df)
    x_main$estimate <- list(numeric(0))
    x_main$std_error <- list(numeric(0))
    if (length(xs_level[[x]][-1]) == 1) {
      x_main$pvalue <- list(numeric(0))
      x_main$df <- list(numeric(0))
    }
    x_main$is_variable_summary <- TRUE
    x_main$is_term_summary <- FALSE
    x_stats <- rbind(x_main, x_stats)
  }
  x_stats$variable <- x
  x_stats$variable_label <- if (inherits(fit_glm, "glm")) {
    formatters::var_labels(fit_glm$data[x], fill = TRUE)
  } else {
    x
  }
  x_stats$interaction <- ""
  x_stats$interaction_label <- ""
  x_stats$reference <- ""
  x_stats$reference_label <- ""
  rownames(x_stats) <- NULL
  x_stats[c(
    "variable",
    "variable_label",
    "term",
    "term_label",
    "interaction",
    "interaction_label",
    "reference",
    "reference_label",
    "estimate",
    "std_error",
    "df",
    "pvalue",
    "is_variable_summary",
    "is_term_summary"
  )]
}

#' @describeIn h_logistic_regression Helper function to tabulate the interaction term
#'   results of a logistic regression model.
#'
#' @return Tabulated interaction term results from a logistic regression model.
#'
#' @examples
#' h_glm_interaction_extract("ARMCD:AGE", mod2)
#'
#' @export
h_glm_interaction_extract <- function(x, fit_glm) {
  vars <- h_get_interaction_vars(fit_glm)
  xs_class <- attr(fit_glm$terms, "dataClasses")

  checkmate::assert_string(x)

  # Only take two-way interaction
  checkmate::assert_vector(vars, len = 2)

  # Only consider simple case: first variable in interaction is arm, a categorical variable
  checkmate::assert_disjunct(xs_class[vars[1]], "numeric")

  xs_level <- fit_glm$xlevels
  xs_coef <- summary(fit_glm)$coefficients
  main_effects <- car::Anova(fit_glm, type = 3, test.statistic = "Wald")
  stats <- c("estimate" = "Estimate", "std_error" = "Std. Error", "pvalue" = "Pr(>|z|)")
  v1_comp <- xs_level[[vars[1]]][-1]
  if (xs_class[vars[2]] == "numeric") {
    x_stats <- as.data.frame(
      xs_coef[paste0(vars[1], v1_comp, ":", vars[2]), stats, drop = FALSE],
      stringsAsFactors = FALSE
    )
    colnames(x_stats) <- names(stats)
    x_stats$term <- v1_comp
    x_numbers <- table(fit_glm$data[[vars[1]]])
    x_stats$term_label <- h_simple_term_labels(v1_comp, x_numbers)
    v1_ref <- xs_level[[vars[1]]][1]
    term_main <- v1_ref
    ref_label <- h_simple_term_labels(v1_ref, x_numbers)
  } else if (xs_class[vars[2]] != "numeric") {
    v2_comp <- xs_level[[vars[2]]][-1]
    v1_v2_grid <- expand.grid(v1 = v1_comp, v2 = v2_comp)
    x_sel <- paste(
      paste0(vars[1], v1_v2_grid$v1),
      paste0(vars[2], v1_v2_grid$v2),
      sep = ":"
    )
    x_stats <- as.data.frame(xs_coef[x_sel, stats, drop = FALSE], stringsAsFactors = FALSE)
    colnames(x_stats) <- names(stats)
    x_stats$term <- paste(v1_v2_grid$v1, "*", v1_v2_grid$v2)
    x_numbers <- table(fit_glm$data[[vars[1]]], fit_glm$data[[vars[2]]])
    x_stats$term_label <- h_interaction_term_labels(v1_v2_grid$v1, v1_v2_grid$v2, x_numbers)
    v1_ref <- xs_level[[vars[1]]][1]
    v2_ref <- xs_level[[vars[2]]][1]
    term_main <- paste(vars[1], vars[2], sep = " * ")
    ref_label <- h_interaction_term_labels(v1_ref, v2_ref, x_numbers, any = TRUE)
  }
  x_stats$df <- as.list(1)
  x_stats$pvalue <- as.list(x_stats$pvalue)
  x_stats$is_variable_summary <- FALSE
  x_stats$is_term_summary <- TRUE
  x_main <- data.frame(
    pvalue = main_effects[x, "Pr(>Chisq)", drop = TRUE],
    term = term_main,
    term_label = paste("Reference", ref_label),
    df = main_effects[x, "Df", drop = TRUE],
    stringsAsFactors = FALSE
  )
  x_main$pvalue <- as.list(x_main$pvalue)
  x_main$df <- as.list(x_main$df)
  x_main$estimate <- list(numeric(0))
  x_main$std_error <- list(numeric(0))
  x_main$is_variable_summary <- TRUE
  x_main$is_term_summary <- FALSE

  x_stats <- rbind(x_main, x_stats)
  x_stats$variable <- x
  x_stats$variable_label <- paste(
    "Interaction of",
    formatters::var_labels(fit_glm$data[vars[1]], fill = TRUE),
    "*",
    formatters::var_labels(fit_glm$data[vars[2]], fill = TRUE)
  )
  x_stats$interaction <- ""
  x_stats$interaction_label <- ""
  x_stats$reference <- ""
  x_stats$reference_label <- ""
  rownames(x_stats) <- NULL
  x_stats[c(
    "variable",
    "variable_label",
    "term",
    "term_label",
    "interaction",
    "interaction_label",
    "reference",
    "reference_label",
    "estimate",
    "std_error",
    "df",
    "pvalue",
    "is_variable_summary",
    "is_term_summary"
  )]
}

#' @describeIn h_logistic_regression Helper function to tabulate the interaction
#'   results of a logistic regression model. This basically is a wrapper for
#'   [h_or_interaction()] and [h_glm_simple_term_extract()] which puts the results
#'   in the right data frame format.
#'
#' @return A `data.frame` of tabulated interaction term results from a logistic regression model.
#'
#' @examples
#' h_glm_inter_term_extract("AGE", "ARMCD", mod2)
#'
#' @export
h_glm_inter_term_extract <- function(odds_ratio_var,
                                     interaction_var,
                                     fit_glm,
                                     ...) {
  # First obtain the main effects.
  main_stats <- h_glm_simple_term_extract(odds_ratio_var, fit_glm)
  main_stats$is_reference_summary <- FALSE
  main_stats$odds_ratio <- NA
  main_stats$lcl <- NA
  main_stats$ucl <- NA

  # Then we get the odds ratio estimates and put into df form.
  or_numbers <- h_or_interaction(odds_ratio_var, interaction_var, fit_glm, ...)
  is_num_or_var <- attr(fit_glm$terms, "dataClasses")[odds_ratio_var] == "numeric"

  if (is_num_or_var) {
    # Numeric OR variable case.
    references <- names(or_numbers)
    n_ref <- length(references)

    extract_from_list <- function(l, name, pos = 1) {
      unname(unlist(
        lapply(or_numbers, function(x) {
          x[[name]][pos]
        })
      ))
    }
    or_stats <- data.frame(
      variable = odds_ratio_var,
      variable_label = unname(formatters::var_labels(fit_glm$data[odds_ratio_var], fill = TRUE)),
      term = odds_ratio_var,
      term_label = unname(formatters::var_labels(fit_glm$data[odds_ratio_var], fill = TRUE)),
      interaction = interaction_var,
      interaction_label = unname(formatters::var_labels(fit_glm$data[interaction_var], fill = TRUE)),
      reference = references,
      reference_label = references,
      estimate = NA,
      std_error = NA,
      odds_ratio = extract_from_list(or_numbers, "or"),
      lcl = extract_from_list(or_numbers, "ci", pos = "lcl"),
      ucl = extract_from_list(or_numbers, "ci", pos = "ucl"),
      df = NA,
      pvalue = NA,
      is_variable_summary = FALSE,
      is_term_summary = FALSE,
      is_reference_summary = TRUE
    )
  } else {
    # Categorical OR variable case.
    references <- names(or_numbers[[1]])
    n_ref <- length(references)

    extract_from_list <- function(l, name, pos = 1) {
      unname(unlist(
        lapply(or_numbers, function(x) {
          lapply(x, function(y) y[[name]][pos])
        })
      ))
    }
    or_stats <- data.frame(
      variable = odds_ratio_var,
      variable_label = unname(formatters::var_labels(fit_glm$data[odds_ratio_var], fill = TRUE)),
      term = rep(names(or_numbers), each = n_ref),
      term_label = h_simple_term_labels(rep(names(or_numbers), each = n_ref), table(fit_glm$data[[odds_ratio_var]])),
      interaction = interaction_var,
      interaction_label = unname(formatters::var_labels(fit_glm$data[interaction_var], fill = TRUE)),
      reference = unlist(lapply(or_numbers, names)),
      reference_label = unlist(lapply(or_numbers, names)),
      estimate = NA,
      std_error = NA,
      odds_ratio = extract_from_list(or_numbers, "or"),
      lcl = extract_from_list(or_numbers, "ci", pos = "lcl"),
      ucl = extract_from_list(or_numbers, "ci", pos = "ucl"),
      df = NA,
      pvalue = NA,
      is_variable_summary = FALSE,
      is_term_summary = FALSE,
      is_reference_summary = TRUE
    )
  }

  df <- rbind(
    main_stats[, names(or_stats)],
    or_stats
  )
  df[order(-df$is_variable_summary, df$term, -df$is_term_summary, df$reference), ]
}

#' @describeIn h_logistic_regression Helper function to tabulate the results including
#'   odds ratios and confidence intervals of simple terms.
#'
#' @return Tabulated statistics for the given variable(s) from the logistic regression model.
#'
#' @examples
#' h_logistic_simple_terms("AGE", mod1)
#'
#' @export
h_logistic_simple_terms <- function(x, fit_glm, conf_level = 0.95) {
  checkmate::assert_multi_class(fit_glm, c("glm", "clogit"))
  if (inherits(fit_glm, "glm")) {
    checkmate::assert_set_equal(fit_glm$family$family, "binomial")
  }
  terms_name <- attr(stats::terms(fit_glm), "term.labels")
  xs_class <- attr(fit_glm$terms, "dataClasses")
  interaction <- terms_name[which(!terms_name %in% names(xs_class))]
  checkmate::assert_subset(x, terms_name)
  if (length(interaction) != 0) {
    # Make sure any item in x is not part of interaction term
    checkmate::assert_disjunct(x, unlist(strsplit(interaction, ":")))
  }
  x_stats <- lapply(x, h_glm_simple_term_extract, fit_glm)
  x_stats <- do.call(rbind, x_stats)
  q_norm <- stats::qnorm((1 + conf_level) / 2)
  x_stats$odds_ratio <- lapply(x_stats$estimate, exp)
  x_stats$lcl <- Map(function(or, se) exp(log(or) - q_norm * se), x_stats$odds_ratio, x_stats$std_error)
  x_stats$ucl <- Map(function(or, se) exp(log(or) + q_norm * se), x_stats$odds_ratio, x_stats$std_error)
  x_stats$ci <- Map(function(lcl, ucl) c(lcl, ucl), lcl = x_stats$lcl, ucl = x_stats$ucl)
  x_stats
}

#' @describeIn h_logistic_regression Helper function to tabulate the results including
#'   odds ratios and confidence intervals of interaction terms.
#'
#' @return Tabulated statistics for the given variable(s) from the logistic regression model.
#'
#' @examples
#' h_logistic_inter_terms(c("RACE", "AGE", "ARMCD", "AGE:ARMCD"), mod2)
#'
#' @export
h_logistic_inter_terms <- function(x,
                                   fit_glm,
                                   conf_level = 0.95,
                                   at = NULL) {
  # Find out the interaction variables and interaction term.
  inter_vars <- h_get_interaction_vars(fit_glm)
  checkmate::assert_vector(inter_vars, len = 2)


  inter_term_index <- intersect(grep(inter_vars[1], x), grep(inter_vars[2], x))
  inter_term <- x[inter_term_index]

  # For the non-interaction vars we need the standard stuff.
  normal_terms <- setdiff(x, union(inter_vars, inter_term))

  x_stats <- lapply(normal_terms, h_glm_simple_term_extract, fit_glm)
  x_stats <- do.call(rbind, x_stats)
  q_norm <- stats::qnorm((1 + conf_level) / 2)
  x_stats$odds_ratio <- lapply(x_stats$estimate, exp)
  x_stats$lcl <- Map(function(or, se) exp(log(or) - q_norm * se), x_stats$odds_ratio, x_stats$std_error)
  x_stats$ucl <- Map(function(or, se) exp(log(or) + q_norm * se), x_stats$odds_ratio, x_stats$std_error)
  normal_stats <- x_stats
  normal_stats$is_reference_summary <- FALSE

  # Now the interaction term itself.
  inter_term_stats <- h_glm_interaction_extract(inter_term, fit_glm)
  inter_term_stats$odds_ratio <- NA
  inter_term_stats$lcl <- NA
  inter_term_stats$ucl <- NA
  inter_term_stats$is_reference_summary <- FALSE

  is_intervar1_numeric <- attr(fit_glm$terms, "dataClasses")[inter_vars[1]] == "numeric"

  # Interaction stuff.
  inter_stats_one <- h_glm_inter_term_extract(
    inter_vars[1],
    inter_vars[2],
    fit_glm,
    conf_level = conf_level,
    at = `if`(is_intervar1_numeric, NULL, at)
  )
  inter_stats_two <- h_glm_inter_term_extract(
    inter_vars[2],
    inter_vars[1],
    fit_glm,
    conf_level = conf_level,
    at = `if`(is_intervar1_numeric, at, NULL)
  )

  # Now just combine everything in one data frame.
  col_names <- c(
    "variable",
    "variable_label",
    "term",
    "term_label",
    "interaction",
    "interaction_label",
    "reference",
    "reference_label",
    "estimate",
    "std_error",
    "df",
    "pvalue",
    "odds_ratio",
    "lcl",
    "ucl",
    "is_variable_summary",
    "is_term_summary",
    "is_reference_summary"
  )
  df <- rbind(
    inter_stats_one[, col_names],
    inter_stats_two[, col_names],
    inter_term_stats[, col_names]
  )
  if (length(normal_terms) > 0) {
    df <- rbind(
      normal_stats[, col_names],
      df
    )
  }
  df$ci <- combine_vectors(df$lcl, df$ucl)
  df
}

#' Create a forest plot from an `rtable`
#'
#' Given a [rtables::rtable()] object with at least one column with a single value and one column with 2
#' values, converts table to a [ggplot2::ggplot()] object and generates an accompanying forest plot. The
#' table and forest plot are printed side-by-side.
#'
#' @description `r lifecycle::badge("stable")`
#'
#' @inheritParams rtable2gg
#' @inheritParams argument_convention
#' @param tbl (`VTableTree`)\cr `rtables` table with at least one column with a single value and one column with 2
#'   values.
#' @param col_x (`integer(1)` or `NULL`)\cr column index with estimator. By default tries to get this from
#'   `tbl` attribute `col_x`, otherwise needs to be manually specified. If `NULL`, points will be excluded
#'   from forest plot.
#' @param col_ci (`integer(1)` or `NULL`)\cr column index with confidence intervals. By default tries to get this from
#'   `tbl` attribute `col_ci`, otherwise needs to be manually specified. If `NULL`, lines will be excluded
#'   from forest plot.
#' @param vline (`numeric(1)` or `NULL`)\cr x coordinate for vertical line, if `NULL` then the line is omitted.
#' @param forest_header (`character(2)`)\cr text displayed to the left and right of `vline`, respectively.
#'   If `vline = NULL` then `forest_header` is not printed. By default tries to get this from `tbl` attribute
#'   `forest_header`. If `NULL`, defaults will be extracted from the table if possible, and set to
#'   `"Comparison\nBetter"` and `"Treatment\nBetter"` if not.
#' @param xlim (`numeric(2)`)\cr limits for x axis.
#' @param logx (`flag`)\cr show the x-values on logarithm scale.
#' @param x_at (`numeric`)\cr x-tick locations, if `NULL`, `x_at` is set to `vline` and both `xlim` values.
#' @param width_row_names `r lifecycle::badge("deprecated")` Please use the `lbl_col_padding` argument instead.
#' @param width_columns (`numeric`)\cr a vector of column widths. Each element's position in
#'   `colwidths` corresponds to the column of `tbl` in the same position. If `NULL`, column widths are calculated
#'   according to maximum number of characters per column.
#' @param width_forest `r lifecycle::badge("deprecated")` Please use the `rel_width_forest` argument instead.
#' @param rel_width_forest (`proportion`)\cr proportion of total width to allocate to the forest plot. Relative
#'   width of table is then `1 - rel_width_forest`. If `as_list = TRUE`, this parameter is ignored.
#' @param font_size (`numeric(1)`)\cr font size.
#' @param col_symbol_size (`numeric` or `NULL`)\cr column index from `tbl` containing data to be used
#'   to determine relative size for estimator plot symbol. Typically, the symbol size is proportional
#'   to the sample size used to calculate the estimator. If `NULL`, the same symbol size is used for all subgroups.
#'   By default tries to get this from `tbl` attribute `col_symbol_size`, otherwise needs to be manually specified.
#' @param col (`character`)\cr color(s).
#' @param ggtheme (`theme`)\cr a graphical theme as provided by `ggplot2` to control styling of the plot.
#' @param as_list (`flag`)\cr whether the two `ggplot` objects should be returned as a list. If `TRUE`, a named list
#'   with two elements, `table` and `plot`, will be returned. If `FALSE` (default) the table and forest plot are
#'   printed side-by-side via [cowplot::plot_grid()].
#' @param gp `r lifecycle::badge("deprecated")` `g_forest` is now generated as a `ggplot` object. This argument
#'   is no longer used.
#' @param draw `r lifecycle::badge("deprecated")` `g_forest` is now generated as a `ggplot` object. This argument
#'   is no longer used.
#' @param newpage `r lifecycle::badge("deprecated")` `g_forest` is now generated as a `ggplot` object. This argument
#'   is no longer used.
#'
#' @return `ggplot` forest plot and table.
#'
#' @examples
#' library(dplyr)
#' library(forcats)
#' library(nestcolor)
#'
#' adrs <- tern_ex_adrs
#' n_records <- 20
#' adrs_labels <- formatters::var_labels(adrs, fill = TRUE)
#' adrs <- adrs %>%
#'   filter(PARAMCD == "BESRSPI") %>%
#'   filter(ARM %in% c("A: Drug X", "B: Placebo")) %>%
#'   slice(seq_len(n_records)) %>%
#'   droplevels() %>%
#'   mutate(
#'     # Reorder levels of factor to make the placebo group the reference arm.
#'     ARM = fct_relevel(ARM, "B: Placebo"),
#'     rsp = AVALC == "CR"
#'   )
#' formatters::var_labels(adrs) <- c(adrs_labels, "Response")
#' df <- extract_rsp_subgroups(
#'   variables = list(rsp = "rsp", arm = "ARM", subgroups = c("SEX", "STRATA2")),
#'   data = adrs
#' )
#' # Full commonly used response table.
#'
#' tbl <- basic_table() %>%
#'   tabulate_rsp_subgroups(df)
#' g_forest(tbl)
#'
#' # Odds ratio only table.
#'
#' tbl_or <- basic_table() %>%
#'   tabulate_rsp_subgroups(df, vars = c("n_tot", "or", "ci"))
#' g_forest(
#'   tbl_or,
#'   forest_header = c("Comparison\nBetter", "Treatment\nBetter")
#' )
#'
#' # Survival forest plot example.
#' adtte <- tern_ex_adtte
#' # Save variable labels before data processing steps.
#' adtte_labels <- formatters::var_labels(adtte, fill = TRUE)
#' adtte_f <- adtte %>%
#'   filter(
#'     PARAMCD == "OS",
#'     ARM %in% c("B: Placebo", "A: Drug X"),
#'     SEX %in% c("M", "F")
#'   ) %>%
#'   mutate(
#'     # Reorder levels of ARM to display reference arm before treatment arm.
#'     ARM = droplevels(fct_relevel(ARM, "B: Placebo")),
#'     SEX = droplevels(SEX),
#'     AVALU = as.character(AVALU),
#'     is_event = CNSR == 0
#'   )
#' labels <- list(
#'   "ARM" = adtte_labels["ARM"],
#'   "SEX" = adtte_labels["SEX"],
#'   "AVALU" = adtte_labels["AVALU"],
#'   "is_event" = "Event Flag"
#' )
#' formatters::var_labels(adtte_f)[names(labels)] <- as.character(labels)
#' df <- extract_survival_subgroups(
#'   variables = list(
#'     tte = "AVAL",
#'     is_event = "is_event",
#'     arm = "ARM", subgroups = c("SEX", "BMRKR2")
#'   ),
#'   data = adtte_f
#' )
#' table_hr <- basic_table() %>%
#'   tabulate_survival_subgroups(df, time_unit = adtte_f$AVALU[1])
#' g_forest(table_hr)
#'
#' # Works with any `rtable`.
#' tbl <- rtable(
#'   header = c("E", "CI", "N"),
#'   rrow("", 1, c(.8, 1.2), 200),
#'   rrow("", 1.2, c(1.1, 1.4), 50)
#' )
#' g_forest(
#'   tbl = tbl,
#'   col_x = 1,
#'   col_ci = 2,
#'   xlim = c(0.5, 2),
#'   x_at = c(0.5, 1, 2),
#'   col_symbol_size = 3
#' )
#'
#' tbl <- rtable(
#'   header = rheader(
#'     rrow("", rcell("A", colspan = 2)),
#'     rrow("", "c1", "c2")
#'   ),
#'   rrow("row 1", 1, c(.8, 1.2)),
#'   rrow("row 2", 1.2, c(1.1, 1.4))
#' )
#' g_forest(
#'   tbl = tbl,
#'   col_x = 1,
#'   col_ci = 2,
#'   xlim = c(0.5, 2),
#'   x_at = c(0.5, 1, 2),
#'   vline = 1,
#'   forest_header = c("Hello", "World")
#' )
#'
#' @export
g_forest <- function(tbl,
                     col_x = attr(tbl, "col_x"),
                     col_ci = attr(tbl, "col_ci"),
                     vline = 1,
                     forest_header = attr(tbl, "forest_header"),
                     xlim = c(0.1, 10),
                     logx = TRUE,
                     x_at = c(0.1, 1, 10),
                     width_row_names = lifecycle::deprecated(),
                     width_columns = NULL,
                     width_forest = lifecycle::deprecated(),
                     lbl_col_padding = 0,
                     rel_width_forest = 0.25,
                     font_size = 12,
                     col_symbol_size = attr(tbl, "col_symbol_size"),
                     col = getOption("ggplot2.discrete.colour")[1],
                     ggtheme = NULL,
                     as_list = FALSE,
                     gp = lifecycle::deprecated(),
                     draw = lifecycle::deprecated(),
                     newpage = lifecycle::deprecated()) {
  # Deprecated argument warnings
  if (lifecycle::is_present(width_row_names)) {
    lifecycle::deprecate_warn(
      "0.9.4", "g_forest(width_row_names)", "g_forest(lbl_col_padding)",
      details = "The width of the row label column can be adjusted via the `lbl_col_padding` parameter."
    )
  }
  if (lifecycle::is_present(width_forest)) {
    lifecycle::deprecate_warn(
      "0.9.4", "g_forest(width_forest)", "g_forest(rel_width_forest)",
      details = "Relative width of the forest plot (as a proportion) can be set via the `rel_width_forest` parameter."
    )
  }
  if (lifecycle::is_present(gp)) {
    lifecycle::deprecate_warn(
      "0.9.4", "g_forest(gp)", "g_forest(ggtheme)",
      details = paste(
        "`g_forest` is now generated as a `ggplot` object.",
        "Additional display settings should be supplied via the `ggtheme` parameter."
      )
    )
  }
  if (lifecycle::is_present(draw)) {
    lifecycle::deprecate_warn(
      "0.9.4", "g_forest(draw)",
      details = "`g_forest` now generates `ggplot` objects. This parameter has no effect."
    )
  }
  if (lifecycle::is_present(newpage)) {
    lifecycle::deprecate_warn(
      "0.9.4", "g_forest(newpage)",
      details = "`g_forest` now generates `ggplot` objects. This parameter has no effect."
    )
  }

  checkmate::assert_class(tbl, "VTableTree")
  checkmate::assert_number(col_x, lower = 0, upper = ncol(tbl), null.ok = TRUE)
  checkmate::assert_number(col_ci, lower = 0, upper = ncol(tbl), null.ok = TRUE)
  checkmate::assert_number(col_symbol_size, lower = 0, upper = ncol(tbl), null.ok = TRUE)
  checkmate::assert_number(font_size, lower = 0)
  checkmate::assert_character(col, null.ok = TRUE)
  checkmate::assert_true(is.null(col) | length(col) == 1 | length(col) == nrow(tbl))

  # Extract info from table
  mat <- matrix_form(tbl, indent_rownames = TRUE)
  mat_strings <- formatters::mf_strings(mat)
  nlines_hdr <- formatters::mf_nlheader(mat)
  nrows_body <- nrow(mat_strings) - nlines_hdr
  tbl_stats <- mat_strings[nlines_hdr, -1]

  # Generate and modify table as ggplot object
  gg_table <- rtable2gg(tbl, fontsize = font_size, colwidths = width_columns, lbl_col_padding = lbl_col_padding) +
    theme(plot.margin = margin(0, 0, 0, 0.025, "npc"))
  gg_table$scales$scales[[1]]$expand <- c(0.01, 0.01)
  gg_table$scales$scales[[2]]$limits[2] <- nrow(mat_strings) + 1
  if (nlines_hdr == 2) {
    gg_table$scales$scales[[2]]$expand <- c(0, 0)
    arms <- unique(mat_strings[1, ][nzchar(trimws(mat_strings[1, ]))])
  } else {
    arms <- NULL
  }

  tbl_df <- as_result_df(tbl)
  dat_cols <- seq(which(names(tbl_df) == "node_class") + 1, ncol(tbl_df))
  tbl_df <- tbl_df[, c(which(names(tbl_df) == "row_num"), dat_cols)]
  names(tbl_df) <- c("row_num", tbl_stats)

  # Check table data columns
  if (!is.null(col_ci)) {
    ci_col <- col_ci + 1
  } else {
    tbl_df[["empty_ci"]] <- rep(list(c(NA_real_, NA_real_)), nrow(tbl_df))
    ci_col <- which(names(tbl_df) == "empty_ci")
  }
  if (length(tbl_df[, ci_col][[1]]) != 2) stop("CI column must have two elements (lower and upper limits).")

  if (!is.null(col_x)) {
    x_col <- col_x + 1
  } else {
    tbl_df[["empty_x"]] <- NA_real_
    x_col <- which(names(tbl_df) == "empty_x")
  }
  if (!is.null(col_symbol_size)) {
    sym_size <- unlist(tbl_df[, col_symbol_size + 1])
  } else {
    sym_size <- rep(1, nrow(tbl_df))
  }

  tbl_df[, c("ci_lwr", "ci_upr")] <- t(sapply(tbl_df[, ci_col], unlist))
  x <- unlist(tbl_df[, x_col])
  lwr <- unlist(tbl_df[["ci_lwr"]])
  upr <- unlist(tbl_df[["ci_upr"]])
  row_num <- nrow(mat_strings) - tbl_df[["row_num"]] - as.numeric(nlines_hdr == 2)

  if (is.null(col)) col <- "#343cff"
  if (length(col) == 1) col <- rep(col, nrow(tbl_df))
  if (is.null(x_at)) x_at <- union(xlim, vline)
  x_labels <- x_at

  # Apply log transformation
  if (logx) {
    x_t <- log(x)
    lwr_t <- log(lwr)
    upr_t <- log(upr)
    xlim_t <- log(xlim)
  } else {
    x_t <- x
    lwr_t <- lwr
    upr_t <- upr
    xlim_t <- xlim
  }

  # Set up plot area
  gg_plt <- ggplot(data = tbl_df) +
    theme(
      panel.background = element_rect(fill = "transparent", color = NA_character_),
      plot.background = element_rect(fill = "transparent", color = NA_character_),
      panel.grid.major = element_blank(),
      panel.grid.minor = element_blank(),
      axis.title.x = element_blank(),
      axis.title.y = element_blank(),
      axis.line.x = element_line(),
      axis.text = element_text(size = font_size),
      legend.position = "none",
      plot.margin = margin(0, 0.1, 0.05, 0, "npc")
    ) +
    scale_x_continuous(
      transform = ifelse(logx, "log", "identity"),
      limits = xlim,
      breaks = x_at,
      labels = x_labels,
      expand = c(0.01, 0)
    ) +
    scale_y_continuous(
      limits = c(0, nrow(mat_strings) + 1),
      breaks = NULL,
      expand = c(0, 0)
    ) +
    coord_cartesian(clip = "off")

  if (is.null(ggtheme)) {
    gg_plt <- gg_plt + annotate(
      "rect",
      xmin = xlim[1],
      xmax = xlim[2],
      ymin = 0,
      ymax = nrows_body + 0.5,
      fill = "grey92"
    )
  }

  if (!is.null(vline)) {
    # Set default forest header
    if (is.null(forest_header)) {
      forest_header <- c(
        paste(if (length(arms) == 2) arms[1] else "Comparison", "Better", sep = "\n"),
        paste(if (length(arms) == 2) arms[2] else "Treatment", "Better", sep = "\n")
      )
    }

    # Add vline and forest header labels
    mid_pts <- if (logx) {
      c(exp(mean(log(c(xlim[1], vline)))), exp(mean(log(c(vline, xlim[2])))))
    } else {
      c(mean(c(xlim[1], vline)), mean(c(vline, xlim[2])))
    }
    gg_plt <- gg_plt +
      annotate(
        "segment",
        x = vline, xend = vline, y = 0, yend = nrows_body + 0.5
      ) +
      annotate(
        "text",
        x = mid_pts[1], y = nrows_body + 1.25,
        label = forest_header[1],
        size = font_size / .pt,
        lineheight = 0.9
      ) +
      annotate(
        "text",
        x = mid_pts[2], y = nrows_body + 1.25,
        label = forest_header[2],
        size = font_size / .pt,
        lineheight = 0.9
      )
  }

  # Add points to plot
  if (any(!is.na(x_t))) {
    x_t[x < xlim[1] | x > xlim[2]] <- NA
    gg_plt <- gg_plt + geom_point(
      x = x_t,
      y = row_num,
      color = col,
      aes(size = sym_size),
      na.rm = TRUE
    )
  }

  for (i in seq_len(nrow(tbl_df))) {
    # Determine which arrow(s) to add to CI lines
    which_arrow <- c(lwr_t[i] < xlim_t[1], upr_t[i] > xlim_t[2])
    which_arrow <- dplyr::case_when(
      all(which_arrow) ~ "both",
      which_arrow[1] ~ "first",
      which_arrow[2] ~ "last",
      TRUE ~ NA_character_
    )

    # Add CI lines
    gg_plt <- gg_plt +
      if (!is.na(which_arrow)) {
        annotate(
          "segment",
          x = if (!which_arrow %in% c("first", "both")) lwr[i] else xlim[1],
          xend = if (!which_arrow %in% c("last", "both")) upr[i] else xlim[2],
          y = row_num[i], yend = row_num[i],
          color = if (length(col) == 1) col else col[i],
          arrow = arrow(length = unit(0.05, "npc"), ends = which_arrow),
          na.rm = TRUE
        )
      } else {
        annotate(
          "segment",
          x = lwr[i], xend = upr[i],
          y = row_num[i], yend = row_num[i],
          color = if (length(col) == 1) col else col[i],
          na.rm = TRUE
        )
      }
  }

  # Apply custom ggtheme to plot
  if (!is.null(ggtheme)) gg_plt <- gg_plt + ggtheme

  if (as_list) {
    list(
      table = gg_table,
      plot = gg_plt
    )
  } else {
    cowplot::plot_grid(
      gg_table,
      gg_plt,
      align = "h",
      axis = "tblr",
      rel_widths = c(1 - rel_width_forest, rel_width_forest)
    )
  }
}

#' Forest plot grob
#'
#' @description `r lifecycle::badge("deprecated")`
#'
#' @inheritParams g_forest
#' @param tbl (`VTableTree`)\cr `rtables` table object.
#' @param x (`numeric`)\cr coordinate of point.
#' @param lower,upper (`numeric`)\cr lower/upper bound of the confidence interval.
#' @param symbol_size (`numeric`)\cr vector with relative size for plot symbol.
#'   If `NULL`, the same symbol size is used.
#'
#' @details
#' The heights get automatically determined.
#'
#' @examples
#' tbl <- rtable(
#'   header = rheader(
#'     rrow("", "E", rcell("CI", colspan = 2), "N"),
#'     rrow("", "A", "B", "C", "D")
#'   ),
#'   rrow("row 1", 1, 0.8, 1.1, 16),
#'   rrow("row 2", 1.4, 0.8, 1.6, 25),
#'   rrow("row 3", 1.2, 0.8, 1.6, 36)
#' )
#'
#' x <- c(1, 1.4, 1.2)
#' lower <- c(0.8, 0.8, 0.8)
#' upper <- c(1.1, 1.6, 1.6)
#' # numeric vector with multiplication factor to scale each circle radius
#' # default radius is 1/3.5 lines
#' symbol_scale <- c(1, 1.25, 1.5)
#'
#' # Internal function - forest_grob
#' \donttest{
#' p <- forest_grob(tbl, x, lower, upper,
#'   vline = 1, forest_header = c("A", "B"),
#'   x_at = c(.1, 1, 10), xlim = c(0.1, 10), logx = TRUE, symbol_size = symbol_scale,
#'   vp = grid::plotViewport(margins = c(1, 1, 1, 1))
#' )
#'
#' draw_grob(p)
#' }
#'
#' @noRd
#' @keywords internal
forest_grob <- function(tbl,
                        x,
                        lower,
                        upper,
                        vline,
                        forest_header,
                        xlim = NULL,
                        logx = FALSE,
                        x_at = NULL,
                        width_row_names = NULL,
                        width_columns = NULL,
                        width_forest = grid::unit(1, "null"),
                        symbol_size = NULL,
                        col = "blue",
                        name = NULL,
                        gp = NULL,
                        vp = NULL) {
  lifecycle::deprecate_warn(
    "0.9.4", "forest_grob()",
    details = "`g_forest` now generates `ggplot` objects. This function is no longer used within `tern`."
  )

  nr <- nrow(tbl)
  if (is.null(vline)) {
    checkmate::assert_true(is.null(forest_header))
  } else {
    checkmate::assert_number(vline)
    checkmate::assert_character(forest_header, len = 2, null.ok = TRUE)
  }

  checkmate::assert_numeric(x, len = nr)
  checkmate::assert_numeric(lower, len = nr)
  checkmate::assert_numeric(upper, len = nr)
  checkmate::assert_numeric(symbol_size, len = nr, null.ok = TRUE)
  checkmate::assert_character(col)

  if (is.null(symbol_size)) {
    symbol_size <- rep(1, nr)
  }

  if (is.null(xlim)) {
    r <- range(c(x, lower, upper), na.rm = TRUE)
    xlim <- r + c(-0.05, 0.05) * diff(r)
  }

  if (logx) {
    if (is.null(x_at)) {
      x_at <- pretty(log(stats::na.omit(c(x, lower, upper))))
      x_labels <- exp(x_at)
    } else {
      x_labels <- x_at
      x_at <- log(x_at)
    }
    xlim <- log(xlim)
    x <- log(x)
    lower <- log(lower)
    upper <- log(upper)
    if (!is.null(vline)) {
      vline <- log(vline)
    }
  } else {
    x_labels <- TRUE
  }

  data_forest_vp <- grid::dataViewport(xlim, c(0, 1))

  # Get table content as matrix form.
  mf <- matrix_form(tbl)

  # Use `rtables` indent_string eventually.
  mf$strings[, 1] <- paste0(
    strrep("    ", c(rep(0, attr(mf, "nrow_header")), mf$row_info$indent)),
    mf$strings[, 1]
  )

  n_header <- attr(mf, "nrow_header")

  if (any(mf$display[, 1] == FALSE)) stop("row names need to be always displayed")

  # Pre-process the data to be used in lapply and cell_in_rows.
  to_args_for_cell_in_rows_fun <- function(part = c("body", "header"),
                                           underline_colspan = FALSE) {
    part <- match.arg(part)
    if (part == "body") {
      mat_row_indices <- seq_len(nrow(tbl)) + n_header
      row_ind_offset <- -n_header
    } else {
      mat_row_indices <- seq_len(n_header)
      row_ind_offset <- 0
    }

    lapply(mat_row_indices, function(i) {
      disp <- mf$display[i, -1]
      list(
        row_name = mf$strings[i, 1],
        cells = mf$strings[i, -1][disp],
        cell_spans = mf$spans[i, -1][disp],
        row_index = i + row_ind_offset,
        underline_colspan = underline_colspan
      )
    })
  }

  args_header <- to_args_for_cell_in_rows_fun("header", underline_colspan = TRUE)
  args_body <- to_args_for_cell_in_rows_fun("body", underline_colspan = FALSE)

  grid::gTree(
    name = name,
    children = grid::gList(
      grid::gTree(
        children = do.call(grid::gList, lapply(args_header, do.call, what = cell_in_rows)),
        vp = grid::vpPath("vp_table_layout", "vp_header")
      ),
      grid::gTree(
        children = do.call(grid::gList, lapply(args_body, do.call, what = cell_in_rows)),
        vp = grid::vpPath("vp_table_layout", "vp_body")
      ),
      grid::linesGrob(
        grid::unit(c(0, 1), "npc"),
        y = grid::unit(c(.5, .5), "npc"),
        vp = grid::vpPath("vp_table_layout", "vp_spacer")
      ),
      # forest part
      if (is.null(vline)) {
        NULL
      } else {
        grid::gTree(
          children = grid::gList(
            grid::gTree(
              children = grid::gList(
                # this may overflow, to fix, look here
                # https://stackoverflow.com/questions/33623169/add-multi-line-footnote-to-tablegrob-while-using-gridextra-in-r # nolint
                grid::textGrob(
                  forest_header[1],
                  x = grid::unit(vline, "native") - grid::unit(1, "lines"),
                  just = c("right", "center")
                ),
                grid::textGrob(
                  forest_header[2],
                  x = grid::unit(vline, "native") + grid::unit(1, "lines"),
                  just = c("left", "center")
                )
              ),
              vp = grid::vpStack(grid::viewport(layout.pos.col = ncol(tbl) + 2), data_forest_vp)
            )
          ),
          vp = grid::vpPath("vp_table_layout", "vp_header")
        )
      },
      grid::gTree(
        children = grid::gList(
          grid::gTree(
            children = grid::gList(
              grid::rectGrob(gp = grid::gpar(col = "gray90", fill = "gray90")),
              if (is.null(vline)) {
                NULL
              } else {
                grid::linesGrob(
                  x = grid::unit(rep(vline, 2), "native"),
                  y = grid::unit(c(0, 1), "npc"),
                  gp = grid::gpar(lwd = 2),
                  vp = data_forest_vp
                )
              },
              grid::xaxisGrob(at = x_at, label = x_labels, vp = data_forest_vp)
            ),
            vp = grid::viewport(layout.pos.col = ncol(tbl) + 2)
          )
        ),
        vp = grid::vpPath("vp_table_layout", "vp_body")
      ),
      grid::gTree(
        children = do.call(
          grid::gList,
          Map(
            function(xi, li, ui, row_index, size_i, col) {
              forest_dot_line(
                xi,
                li,
                ui,
                row_index,
                xlim,
                symbol_size = size_i,
                col = col,
                datavp = data_forest_vp
              )
            },
            x,
            lower,
            upper,
            seq_along(x),
            symbol_size,
            col,
            USE.NAMES = FALSE
          )
        ),
        vp = grid::vpPath("vp_table_layout", "vp_body")
      )
    ),
    childrenvp = forest_viewport(tbl, width_row_names, width_columns, width_forest),
    vp = vp,
    gp = gp
  )
}

cell_in_rows <- function(row_name,
                         cells,
                         cell_spans,
                         row_index,
                         underline_colspan = FALSE) {
  checkmate::assert_string(row_name)
  checkmate::assert_character(cells, min.len = 1, any.missing = FALSE)
  checkmate::assert_numeric(cell_spans, len = length(cells), any.missing = FALSE)
  checkmate::assert_number(row_index)
  checkmate::assert_flag(underline_colspan)

  vp_name_rn <- paste0("rowname-", row_index)
  g_rowname <- if (!is.null(row_name) && row_name != "") {
    grid::textGrob(
      name = vp_name_rn,
      label = row_name,
      x = grid::unit(0, "npc"),
      just = c("left", "center"),
      vp = grid::vpPath(paste0("rowname-", row_index))
    )
  } else {
    NULL
  }

  gl_cols <- if (!(length(cells) > 0)) {
    list(NULL)
  } else {
    j <- 1 # column index of cell

    lapply(seq_along(cells), function(k) {
      cell_ascii <- cells[[k]]
      cs <- cell_spans[[k]]

      if (is.na(cell_ascii) || is.null(cell_ascii)) {
        cell_ascii <- "NA"
      }

      cell_name <- paste0("g-cell-", row_index, "-", j)

      cell_grobs <- if (identical(cell_ascii, "")) {
        NULL
      } else {
        if (cs == 1) {
          grid::textGrob(
            label = cell_ascii,
            name = cell_name,
            vp = grid::vpPath(paste0("cell-", row_index, "-", j))
          )
        } else {
          # +1 because of rowname
          vp_joined_cols <- grid::viewport(layout.pos.row = row_index, layout.pos.col = seq(j + 1, j + cs))

          lab <- grid::textGrob(
            label = cell_ascii,
            name = cell_name,
            vp = vp_joined_cols
          )

          if (!underline_colspan || grepl("^[[:space:]]*$", cell_ascii)) {
            lab
          } else {
            grid::gList(
              lab,
              grid::linesGrob(
                x = grid::unit.c(grid::unit(.2, "lines"), grid::unit(1, "npc") - grid::unit(.2, "lines")),
                y = grid::unit(c(0, 0), "npc"),
                vp = vp_joined_cols
              )
            )
          }
        }
      }
      j <<- j + cs

      cell_grobs
    })
  }

  grid::gList(
    g_rowname,
    do.call(grid::gList, gl_cols)
  )
}

#' Graphic object: forest dot line
#'
#' @description `r lifecycle::badge("deprecated")`
#'
#' Calculate the `grob` corresponding to the dot line within the forest plot.
#'
#' @noRd
#' @keywords internal
forest_dot_line <- function(x,
                            lower,
                            upper,
                            row_index,
                            xlim,
                            symbol_size = 1,
                            col = "blue",
                            datavp) {
  lifecycle::deprecate_warn(
    "0.9.4", "forest_dot_line()",
    details = "`g_forest` now generates `ggplot` objects. This function is no longer used within `tern`."
  )

  ci <- c(lower, upper)
  if (any(!is.na(c(x, ci)))) {
    # line
    y <- grid::unit(c(0.5, 0.5), "npc")

    g_line <- if (all(!is.na(ci)) && ci[2] > xlim[1] && ci[1] < xlim[2]) {
      # -
      if (ci[1] >= xlim[1] && ci[2] <= xlim[2]) {
        grid::linesGrob(x = grid::unit(c(ci[1], ci[2]), "native"), y = y)
      } else if (ci[1] < xlim[1] && ci[2] > xlim[2]) {
        # <->
        grid::linesGrob(
          x = grid::unit(xlim, "native"),
          y = y,
          arrow = grid::arrow(angle = 30, length = grid::unit(0.5, "lines"), ends = "both")
        )
      } else if (ci[1] < xlim[1] && ci[2] <= xlim[2]) {
        # <-
        grid::linesGrob(
          x = grid::unit(c(xlim[1], ci[2]), "native"),
          y = y,
          arrow = grid::arrow(angle = 30, length = grid::unit(0.5, "lines"), ends = "first")
        )
      } else if (ci[1] >= xlim[1] && ci[2] > xlim[2]) {
        # ->
        grid::linesGrob(
          x = grid::unit(c(ci[1], xlim[2]), "native"),
          y = y,
          arrow = grid::arrow(angle = 30, length = grid::unit(0.5, "lines"), ends = "last")
        )
      }
    } else {
      NULL
    }

    g_circle <- if (!is.na(x) && x >= xlim[1] && x <= xlim[2]) {
      grid::circleGrob(
        x = grid::unit(x, "native"),
        y = y,
        r = grid::unit(1 / 3.5 * symbol_size, "lines"),
        name = "point"
      )
    } else {
      NULL
    }

    grid::gTree(
      children = grid::gList(
        grid::gTree(
          children = grid::gList(
            grid::gList(
              g_line,
              g_circle
            )
          ),
          vp = datavp,
          gp = grid::gpar(col = col, fill = col)
        )
      ),
      vp = grid::vpPath(paste0("forest-", row_index))
    )
  } else {
    NULL
  }
}

#' Create a viewport tree for the forest plot
#'
#' @description `r lifecycle::badge("deprecated")`
#'
#' @param tbl (`VTableTree`)\cr `rtables` table object.
#' @param width_row_names (`grid::unit`)\cr width of row names.
#' @param width_columns (`grid::unit`)\cr width of column spans.
#' @param width_forest (`grid::unit`)\cr width of the forest plot.
#' @param gap_column (`grid::unit`)\cr gap width between the columns.
#' @param gap_header (`grid::unit`)\cr gap width between the header.
#' @param mat_form (`MatrixPrintForm`)\cr matrix print form of the table.
#'
#' @return A viewport tree.
#'
#' @examples
#' library(grid)
#'
#' tbl <- rtable(
#'   header = rheader(
#'     rrow("", "E", rcell("CI", colspan = 2)),
#'     rrow("", "A", "B", "C")
#'   ),
#'   rrow("row 1", 1, 0.8, 1.1),
#'   rrow("row 2", 1.4, 0.8, 1.6),
#'   rrow("row 3", 1.2, 0.8, 1.2)
#' )
#'
#' \donttest{
#' v <- forest_viewport(tbl)
#'
#' grid::grid.newpage()
#' showViewport(v)
#' }
#'
#' @export
forest_viewport <- function(tbl,
                            width_row_names = NULL,
                            width_columns = NULL,
                            width_forest = grid::unit(1, "null"),
                            gap_column = grid::unit(1, "lines"),
                            gap_header = grid::unit(1, "lines"),
                            mat_form = NULL) {
  lifecycle::deprecate_warn(
    "0.9.4",
    "forest_viewport()",
    details = "`g_forest` now generates `ggplot` objects. This function is no longer used within `tern`."
  )

  checkmate::assert_class(tbl, "VTableTree")
  checkmate::assert_true(grid::is.unit(width_forest))
  if (!is.null(width_row_names)) {
    checkmate::assert_true(grid::is.unit(width_row_names))
  }
  if (!is.null(width_columns)) {
    checkmate::assert_true(grid::is.unit(width_columns))
  }

  if (is.null(mat_form)) mat_form <- matrix_form(tbl)

  mat_form$strings[!mat_form$display] <- ""

  nr <- nrow(tbl)
  nc <- ncol(tbl)
  nr_h <- attr(mat_form, "nrow_header")

  if (is.null(width_row_names) || is.null(width_columns)) {
    tbl_widths <- formatters::propose_column_widths(mat_form)
    strs_with_width <- strrep("x", tbl_widths) # that works for mono spaced fonts
    if (is.null(width_row_names)) width_row_names <- grid::stringWidth(strs_with_width[1])
    if (is.null(width_columns)) width_columns <- grid::stringWidth(strs_with_width[-1])
  }

  # Widths for row name, cols, forest.
  widths <- grid::unit.c(
    width_row_names + gap_column,
    width_columns + gap_column,
    width_forest
  )

  n_lines_per_row <- apply(
    X = mat_form$strings,
    MARGIN = 1,
    FUN = function(row) {
      tmp <- vapply(
        gregexpr("\n", row, fixed = TRUE),
        attr, numeric(1),
        "match.length"
      ) + 1
      max(c(tmp, 1))
    }
  )

  i_header <- seq_len(nr_h)

  height_body_rows <- grid::unit(n_lines_per_row[-i_header] * 1.2, "lines")
  height_header_rows <- grid::unit(n_lines_per_row[i_header] * 1.2, "lines")

  height_body <- grid::unit(sum(n_lines_per_row[-i_header]) * 1.2, "lines")
  height_header <- grid::unit(sum(n_lines_per_row[i_header]) * 1.2, "lines")

  nc_g <- nc + 2 # number of columns incl. row names and forest

  vp_tbl <- grid::vpTree(
    parent = grid::viewport(
      name = "vp_table_layout",
      layout = grid::grid.layout(
        nrow = 3, ncol = 1,
        heights = grid::unit.c(height_header, gap_header, height_body)
      )
    ),
    children = grid::vpList(
      vp_forest_table_part(nr_h, nc_g, 1, 1, widths, height_header_rows, "vp_header"),
      vp_forest_table_part(nr, nc_g, 3, 1, widths, height_body_rows, "vp_body"),
      grid::viewport(name = "vp_spacer", layout.pos.row = 2, layout.pos.col = 1)
    )
  )
  vp_tbl
}

#' Viewport forest plot: table part
#'
#' @description `r lifecycle::badge("deprecated")`
#'
#' Prepares a viewport for the table included in the forest plot.
#'
#' @noRd
#' @keywords internal
vp_forest_table_part <- function(nrow,
                                 ncol,
                                 l_row,
                                 l_col,
                                 widths,
                                 heights,
                                 name) {
  lifecycle::deprecate_warn(
    "0.9.4", "vp_forest_table_part()",
    details = "`g_forest` now generates `ggplot` objects. This function is no longer used within `tern`."
  )

  grid::vpTree(
    grid::viewport(
      name = name,
      layout.pos.row = l_row,
      layout.pos.col = l_col,
      layout = grid::grid.layout(nrow = nrow, ncol = ncol, widths = widths, heights = heights)
    ),
    children = grid::vpList(
      do.call(
        grid::vpList,
        lapply(
          seq_len(nrow), function(i) {
            grid::viewport(layout.pos.row = i, layout.pos.col = 1, name = paste0("rowname-", i))
          }
        )
      ),
      do.call(
        grid::vpList,
        apply(
          expand.grid(seq_len(nrow), seq_len(ncol - 2)),
          1,
          function(x) {
            i <- x[1]
            j <- x[2]
            grid::viewport(layout.pos.row = i, layout.pos.col = j + 1, name = paste0("cell-", i, "-", j))
          }
        )
      ),
      do.call(
        grid::vpList,
        lapply(
          seq_len(nrow),
          function(i) {
            grid::viewport(layout.pos.row = i, layout.pos.col = ncol, name = paste0("forest-", i))
          }
        )
      )
    )
  )
}

#' Forest rendering
#'
#' @description `r lifecycle::badge("deprecated")`
#'
#' Renders the forest grob.
#'
#' @noRd
#' @keywords internal
grid.forest <- function(...) { # nolint
  lifecycle::deprecate_warn(
    "0.9.4", "grid.forest()",
    details = "`g_forest` now generates `ggplot` objects. This function is no longer used within `tern`."
  )

  grid::grid.draw(forest_grob(...))
}

#' Occurrence table pruning
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Family of constructor and condition functions to flexibly prune occurrence tables.
#' The condition functions always return whether the row result is higher than the threshold.
#' Since they are of class [CombinationFunction()] they can be logically combined with other condition
#' functions.
#'
#' @note Since most table specifications are worded positively, we name our constructor and condition
#'   functions positively, too. However, note that the result of [keep_rows()] says what
#'   should be pruned, to conform with the [rtables::prune_table()] interface.
#'
#' @examples
#' \donttest{
#' tab <- basic_table() %>%
#'   split_cols_by("ARM") %>%
#'   split_rows_by("RACE") %>%
#'   split_rows_by("STRATA1") %>%
#'   summarize_row_groups() %>%
#'   analyze_vars("COUNTRY", .stats = "count_fraction") %>%
#'   build_table(DM)
#' }
#'
#' @name prune_occurrences
NULL

#' @describeIn prune_occurrences Constructor for creating pruning functions based on
#'   a row condition function. This removes all analysis rows (`TableRow`) that should be
#'   pruned, i.e., don't fulfill the row condition. It removes the sub-tree if there are no
#'   children left.
#'
#' @param row_condition (`CombinationFunction`)\cr condition function which works on individual
#'   analysis rows and flags whether these should be kept in the pruned table.
#'
#' @return
#' * `keep_rows()` returns a pruning function that can be used with [rtables::prune_table()]
#'   to prune an `rtables` table.
#'
#' @examples
#' \donttest{
#' # `keep_rows`
#' is_non_empty <- !CombinationFunction(all_zero_or_na)
#' prune_table(tab, keep_rows(is_non_empty))
#' }
#'
#' @export
keep_rows <- function(row_condition) {
  checkmate::assert_function(row_condition)
  function(table_tree) {
    if (inherits(table_tree, "TableRow")) {
      return(!row_condition(table_tree))
    }
    children <- tree_children(table_tree)
    identical(length(children), 0L)
  }
}

#' @describeIn prune_occurrences Constructor for creating pruning functions based on
#'   a condition for the (first) content row in leaf tables. This removes all leaf tables where
#'   the first content row does not fulfill the condition. It does not check individual rows.
#'   It then proceeds recursively by removing the sub tree if there are no children left.
#'
#' @param content_row_condition (`CombinationFunction`)\cr condition function which works on individual
#'   first content rows of leaf tables and flags whether these leaf tables should be kept in the pruned table.
#'
#' @return
#' * `keep_content_rows()` returns a pruning function that checks the condition on the first content
#'   row of leaf tables in the table.
#'
#' @examples
#' # `keep_content_rows`
#' \donttest{
#' more_than_twenty <- has_count_in_cols(atleast = 20L, col_names = names(tab))
#' prune_table(tab, keep_content_rows(more_than_twenty))
#' }
#'
#' @export
keep_content_rows <- function(content_row_condition) {
  checkmate::assert_function(content_row_condition)
  function(table_tree) {
    if (is_leaf_table(table_tree)) {
      content_row <- h_content_first_row(table_tree)
      return(!content_row_condition(content_row))
    }
    if (inherits(table_tree, "DataRow")) {
      return(FALSE)
    }
    children <- tree_children(table_tree)
    identical(length(children), 0L)
  }
}

#' @describeIn prune_occurrences Constructor for creating condition functions on total counts in the specified columns.
#'
#' @param atleast (`numeric(1)`)\cr threshold which should be met in order to keep the row.
#' @param ... arguments for row or column access, see [`rtables_access`]: either `col_names` (`character`) including
#'   the names of the columns which should be used, or alternatively `col_indices` (`integer`) giving the indices
#'   directly instead.
#'
#' @return
#' * `has_count_in_cols()` returns a condition function that sums the counts in the specified column.
#'
#' @examples
#' \donttest{
#' more_than_one <- has_count_in_cols(atleast = 1L, col_names = names(tab))
#' prune_table(tab, keep_rows(more_than_one))
#' }
#'
#' @export
has_count_in_cols <- function(atleast, ...) {
  checkmate::assert_count(atleast)
  CombinationFunction(function(table_row) {
    row_counts <- h_row_counts(table_row, ...)
    total_count <- sum(row_counts)
    total_count >= atleast
  })
}

#' @describeIn prune_occurrences Constructor for creating condition functions on any of the counts in
#'   the specified columns satisfying a threshold.
#'
#' @param atleast (`numeric(1)`)\cr threshold which should be met in order to keep the row.
#'
#' @return
#' * `has_count_in_any_col()` returns a condition function that compares the counts in the
#'   specified columns with the threshold.
#'
#' @examples
#' \donttest{
#' # `has_count_in_any_col`
#' any_more_than_one <- has_count_in_any_col(atleast = 1L, col_names = names(tab))
#' prune_table(tab, keep_rows(any_more_than_one))
#' }
#'
#' @export
has_count_in_any_col <- function(atleast, ...) {
  checkmate::assert_count(atleast)
  CombinationFunction(function(table_row) {
    row_counts <- h_row_counts(table_row, ...)
    any(row_counts >= atleast)
  })
}

#' @describeIn prune_occurrences Constructor for creating condition functions on total fraction in
#'   the specified columns.
#'
#' @return
#' * `has_fraction_in_cols()` returns a condition function that sums the counts in the
#'   specified column, and computes the fraction by dividing by the total column counts.
#'
#' @examples
#' \donttest{
#' # `has_fraction_in_cols`
#' more_than_five_percent <- has_fraction_in_cols(atleast = 0.05, col_names = names(tab))
#' prune_table(tab, keep_rows(more_than_five_percent))
#' }
#'
#' @export
has_fraction_in_cols <- function(atleast, ...) {
  assert_proportion_value(atleast, include_boundaries = TRUE)
  CombinationFunction(function(table_row) {
    row_counts <- h_row_counts(table_row, ...)
    total_count <- sum(row_counts)
    col_counts <- h_col_counts(table_row, ...)
    total_n <- sum(col_counts)
    total_percent <- total_count / total_n
    total_percent >= atleast
  })
}

#' @describeIn prune_occurrences Constructor for creating condition functions on any fraction in
#'   the specified columns.
#'
#' @return
#' * `has_fraction_in_any_col()` returns a condition function that looks at the fractions
#'  in the specified columns and checks whether any of them fulfill the threshold.
#'
#' @examples
#' \donttest{
#' # `has_fraction_in_any_col`
#' any_atleast_five_percent <- has_fraction_in_any_col(atleast = 0.05, col_names = names(tab))
#' prune_table(tab, keep_rows(any_atleast_five_percent))
#' }
#'
#' @export
has_fraction_in_any_col <- function(atleast, ...) {
  assert_proportion_value(atleast, include_boundaries = TRUE)
  CombinationFunction(function(table_row) {
    row_fractions <- h_row_fractions(table_row, ...)
    any(row_fractions >= atleast)
  })
}

#' @describeIn prune_occurrences Constructor for creating condition function that checks the difference
#'   between the fractions reported in each specified column.
#'
#' @return
#' * `has_fractions_difference()` returns a condition function that extracts the fractions of each
#'   specified column, and computes the difference of the minimum and maximum.
#'
#' @examples
#' \donttest{
#' # `has_fractions_difference`
#' more_than_five_percent_diff <- has_fractions_difference(atleast = 0.05, col_names = names(tab))
#' prune_table(tab, keep_rows(more_than_five_percent_diff))
#' }
#'
#' @export
has_fractions_difference <- function(atleast, ...) {
  assert_proportion_value(atleast, include_boundaries = TRUE)
  CombinationFunction(function(table_row) {
    fractions <- h_row_fractions(table_row, ...)
    difference <- diff(range(fractions))
    difference >= atleast
  })
}

#' @describeIn prune_occurrences Constructor for creating condition function that checks the difference
#'   between the counts reported in each specified column.
#'
#' @return
#' * `has_counts_difference()` returns a condition function that extracts the counts of each
#'   specified column, and computes the difference of the minimum and maximum.
#'
#' @examples
#' \donttest{
#' more_than_one_diff <- has_counts_difference(atleast = 1L, col_names = names(tab))
#' prune_table(tab, keep_rows(more_than_one_diff))
#' }
#'
#' @export
has_counts_difference <- function(atleast, ...) {
  checkmate::assert_count(atleast)
  CombinationFunction(function(table_row) {
    counts <- h_row_counts(table_row, ...)
    difference <- diff(range(counts))
    difference >= atleast
  })
}

#' Cox proportional hazards regression
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Fits a Cox regression model and estimates hazard ratio to describe the effect size in a survival analysis.
#'
#' @inheritParams argument_convention
#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("summarize_coxreg")`
#'   to see available statistics for this function.
#'
#' @details Cox models are the most commonly used methods to estimate the magnitude of
#'   the effect in survival analysis. It assumes proportional hazards: the ratio
#'   of the hazards between groups (e.g., two arms) is constant over time.
#'   This ratio is referred to as the "hazard ratio" (HR) and is one of the
#'   most commonly reported metrics to describe the effect size in survival
#'   analysis (NEST Team, 2020).
#'
#' @seealso [fit_coxreg] for relevant fitting functions, [h_cox_regression] for relevant
#'   helper functions, and [tidy_coxreg] for custom tidy methods.
#'
#' @examples
#' library(survival)
#'
#' # Testing dataset [survival::bladder].
#' set.seed(1, kind = "Mersenne-Twister")
#' dta_bladder <- with(
#'   data = bladder[bladder$enum < 5, ],
#'   tibble::tibble(
#'     TIME = stop,
#'     STATUS = event,
#'     ARM = as.factor(rx),
#'     COVAR1 = as.factor(enum) %>% formatters::with_label("A Covariate Label"),
#'     COVAR2 = factor(
#'       sample(as.factor(enum)),
#'       levels = 1:4, labels = c("F", "F", "M", "M")
#'     ) %>% formatters::with_label("Sex (F/M)")
#'   )
#' )
#' dta_bladder$AGE <- sample(20:60, size = nrow(dta_bladder), replace = TRUE)
#' dta_bladder$STUDYID <- factor("X")
#'
#' u1_variables <- list(
#'   time = "TIME", event = "STATUS", arm = "ARM", covariates = c("COVAR1", "COVAR2")
#' )
#'
#' u2_variables <- list(time = "TIME", event = "STATUS", covariates = c("COVAR1", "COVAR2"))
#'
#' m1_variables <- list(
#'   time = "TIME", event = "STATUS", arm = "ARM", covariates = c("COVAR1", "COVAR2")
#' )
#'
#' m2_variables <- list(time = "TIME", event = "STATUS", covariates = c("COVAR1", "COVAR2"))
#'
#' @name cox_regression
#' @order 1
NULL

#' @describeIn cox_regression Statistics function that transforms results tabulated
#'   from [fit_coxreg_univar()] or [fit_coxreg_multivar()] into a list.
#'
#' @param model_df (`data.frame`)\cr contains the resulting model fit from a [fit_coxreg]
#'   function with tidying applied via [broom::tidy()].
#' @param .stats (`character`)\cr the names of statistics to be reported among:
#'   * `n`: number of observations (univariate only)
#'   * `hr`: hazard ratio
#'   * `ci`: confidence interval
#'   * `pval`: p-value of the treatment effect
#'   * `pval_inter`: p-value of the interaction effect between the treatment and the covariate (univariate only)
#' @param .which_vars (`character`)\cr which rows should statistics be returned for from the given model.
#'   Defaults to `"all"`. Other options include `"var_main"` for main effects, `"inter"` for interaction effects,
#'   and `"multi_lvl"` for multivariate model covariate level rows. When `.which_vars` is `"all"`, specific
#'   variables can be selected by specifying `.var_nms`.
#' @param .var_nms (`character`)\cr the `term` value of rows in `df` for which `.stats` should be returned. Typically
#'   this is the name of a variable. If using variable labels, `var` should be a vector of both the desired
#'   variable name and the variable label in that order to see all `.stats` related to that variable. When `.which_vars`
#'   is `"var_main"`, `.var_nms` should be only the variable name.
#'
#' @return
#' * `s_coxreg()` returns the selected statistic for from the Cox regression model for the selected variable(s).
#'
#' @examples
#' # s_coxreg
#'
#' # Univariate
#' univar_model <- fit_coxreg_univar(variables = u1_variables, data = dta_bladder)
#' df1 <- broom::tidy(univar_model)
#'
#' s_coxreg(model_df = df1, .stats = "hr")
#'
#' # Univariate with interactions
#' univar_model_inter <- fit_coxreg_univar(
#'   variables = u1_variables, control = control_coxreg(interaction = TRUE), data = dta_bladder
#' )
#' df1_inter <- broom::tidy(univar_model_inter)
#'
#' s_coxreg(model_df = df1_inter, .stats = "hr", .which_vars = "inter", .var_nms = "COVAR1")
#'
#' # Univariate without treatment arm - only "COVAR2" covariate effects
#' univar_covs_model <- fit_coxreg_univar(variables = u2_variables, data = dta_bladder)
#' df1_covs <- broom::tidy(univar_covs_model)
#'
#' s_coxreg(model_df = df1_covs, .stats = "hr", .var_nms = c("COVAR2", "Sex (F/M)"))
#'
#' # Multivariate.
#' multivar_model <- fit_coxreg_multivar(variables = m1_variables, data = dta_bladder)
#' df2 <- broom::tidy(multivar_model)
#'
#' s_coxreg(model_df = df2, .stats = "pval", .which_vars = "var_main", .var_nms = "COVAR1")
#' s_coxreg(
#'   model_df = df2, .stats = "pval", .which_vars = "multi_lvl",
#'   .var_nms = c("COVAR1", "A Covariate Label")
#' )
#'
#' # Multivariate without treatment arm - only "COVAR1" main effect
#' multivar_covs_model <- fit_coxreg_multivar(variables = m2_variables, data = dta_bladder)
#' df2_covs <- broom::tidy(multivar_covs_model)
#'
#' s_coxreg(model_df = df2_covs, .stats = "hr")
#'
#' @export
s_coxreg <- function(model_df, .stats, .which_vars = "all", .var_nms = NULL) {
  assert_df_with_variables(model_df, list(term = "term", stat = .stats))
  checkmate::assert_multi_class(model_df$term, classes = c("factor", "character"))
  model_df$term <- as.character(model_df$term)
  .var_nms <- .var_nms[!is.na(.var_nms)]

  if (length(.var_nms) > 0) model_df <- model_df[model_df$term %in% .var_nms, ]
  if (.which_vars == "multi_lvl") model_df$term <- tail(.var_nms, 1)

  # We need a list with names corresponding to the stats to display of equal length to the list of stats.
  y <- split(model_df, f = model_df$term, drop = FALSE)
  y <- stats::setNames(y, nm = rep(.stats, length(y)))

  if (.which_vars == "var_main") {
    y <- lapply(y, function(x) x[1, ]) # only main effect
  } else if (.which_vars %in% c("inter", "multi_lvl")) {
    y <- lapply(y, function(x) if (nrow(y[[1]]) > 1) x[-1, ] else x) # exclude main effect
  }

  lapply(
    X = y,
    FUN = function(x) {
      z <- as.list(x[[.stats]])
      stats::setNames(z, nm = x$term_label)
    }
  )
}

#' @describeIn cox_regression Analysis function which is used as `afun` in [rtables::analyze()]
#'   and `cfun` in [rtables::summarize_row_groups()] within `summarize_coxreg()`.
#'
#' @param eff (`flag`)\cr whether treatment effect should be calculated. Defaults to `FALSE`.
#' @param var_main (`flag`)\cr whether main effects should be calculated. Defaults to `FALSE`.
#' @param na_str (`string`)\cr custom string to replace all `NA` values with. Defaults to `""`.
#' @param cache_env (`environment`)\cr an environment object used to cache the regression model in order to
#'   avoid repeatedly fitting the same model for every row in the table. Defaults to `NULL` (no caching).
#' @param varlabels (`list`)\cr a named list corresponds to the names of variables found in data, passed
#'   as a named list and corresponding to time, event, arm, strata, and covariates terms. If arm is missing
#'   from variables, then only Cox model(s) including the covariates will be fitted and the corresponding
#'   effect estimates will be tabulated later.
#'
#' @return
#' * `a_coxreg()` returns formatted [rtables::CellValue()].
#'
#' @examples
#' a_coxreg(
#'   df = dta_bladder,
#'   labelstr = "Label 1",
#'   variables = u1_variables,
#'   .spl_context = list(value = "COVAR1"),
#'   .stats = "n",
#'   .formats = "xx"
#' )
#'
#' a_coxreg(
#'   df = dta_bladder,
#'   labelstr = "",
#'   variables = u1_variables,
#'   .spl_context = list(value = "COVAR2"),
#'   .stats = "pval",
#'   .formats = "xx.xxxx"
#' )
#'
#' @export
a_coxreg <- function(df,
                     labelstr,
                     eff = FALSE,
                     var_main = FALSE,
                     multivar = FALSE,
                     variables,
                     at = list(),
                     control = control_coxreg(),
                     .spl_context,
                     .stats,
                     .formats,
                     .indent_mods = NULL,
                     na_str = "",
                     cache_env = NULL) {
  cov_no_arm <- !multivar && !"arm" %in% names(variables) && control$interaction # special case: univar no arm
  cov <- tail(.spl_context$value, 1) # current variable/covariate
  var_lbl <- formatters::var_labels(df)[cov] # check for df labels
  if (length(labelstr) > 1) {
    labelstr <- if (cov %in% names(labelstr)) labelstr[[cov]] else var_lbl # use df labels if none
  } else if (!is.na(var_lbl) && labelstr == cov && cov %in% variables$covariates) {
    labelstr <- var_lbl
  }
  if (eff || multivar || cov_no_arm) {
    control$interaction <- FALSE
  } else {
    variables$covariates <- cov
    if (var_main) control$interaction <- TRUE
  }

  if (is.null(cache_env[[cov]])) {
    if (!multivar) {
      model <- fit_coxreg_univar(variables = variables, data = df, at = at, control = control) %>% broom::tidy()
    } else {
      model <- fit_coxreg_multivar(variables = variables, data = df, control = control) %>% broom::tidy()
    }
    cache_env[[cov]] <- model
  } else {
    model <- cache_env[[cov]]
  }
  if (!multivar && !var_main) model[, "pval_inter"] <- NA_real_

  if (cov_no_arm || (!cov_no_arm && !"arm" %in% names(variables) && is.numeric(df[[cov]]))) {
    multivar <- TRUE
    if (!cov_no_arm) var_main <- TRUE
  }

  vars_coxreg <- list(which_vars = "all", var_nms = NULL)
  if (eff) {
    if (multivar && !var_main) { # multivar treatment level
      var_lbl_arm <- formatters::var_labels(df)[[variables$arm]]
      vars_coxreg[c("var_nms", "which_vars")] <- list(c(variables$arm, var_lbl_arm), "multi_lvl")
    } else { # treatment effect
      vars_coxreg["var_nms"] <- variables$arm
      if (var_main) vars_coxreg["which_vars"] <- "var_main"
    }
  } else {
    if (!multivar || (multivar && var_main && !is.numeric(df[[cov]]))) { # covariate effect/level
      vars_coxreg[c("var_nms", "which_vars")] <- list(cov, "var_main")
    } else if (multivar) { # multivar covariate level
      vars_coxreg[c("var_nms", "which_vars")] <- list(c(cov, var_lbl), "multi_lvl")
      if (var_main) model[cov, .stats] <- NA_real_
    }
    if (!multivar && !var_main && control$interaction) vars_coxreg["which_vars"] <- "inter" # interaction effect
  }
  var_vals <- s_coxreg(model, .stats, .which_vars = vars_coxreg$which_vars, .var_nms = vars_coxreg$var_nms)[[1]]
  var_names <- if (all(grepl("\\(reference = ", names(var_vals))) && labelstr != tail(.spl_context$value, 1)) {
    paste(c(labelstr, tail(strsplit(names(var_vals), " ")[[1]], 3)), collapse = " ") # "reference" main effect labels
  } else if ((!multivar && !eff && !(!var_main && control$interaction) && nchar(labelstr) > 0) ||
    (multivar && var_main && is.numeric(df[[cov]]))) { # nolint
    labelstr # other main effect labels
  } else if (multivar && !eff && !var_main && is.numeric(df[[cov]])) {
    "All" # multivar numeric covariate
  } else {
    names(var_vals)
  }
  in_rows(
    .list = var_vals, .names = var_names, .labels = var_names, .indent_mods = .indent_mods,
    .formats = stats::setNames(rep(.formats, length(var_names)), var_names),
    .format_na_strs = stats::setNames(rep(na_str, length(var_names)), var_names)
  )
}

#' @describeIn cox_regression Layout-creating function which creates a Cox regression summary table
#'   layout. This function is a wrapper for several `rtables` layouting functions. This function
#'   is a wrapper for [rtables::analyze_colvars()] and [rtables::summarize_row_groups()].
#'
#' @inheritParams fit_coxreg_univar
#' @param multivar (`flag`)\cr whether multivariate Cox regression should run (defaults to `FALSE`), otherwise
#'   univariate Cox regression will run.
#' @param common_var (`string`)\cr the name of a factor variable in the dataset which takes the same value
#'   for all rows. This should be created during pre-processing if no such variable currently exists.
#' @param .section_div (`string` or `NA`)\cr string which should be repeated as a section divider between sections.
#'   Defaults to `NA` for no section divider. If a vector of two strings are given, the first will be used between
#'   treatment and covariate sections and the second between different covariates.
#'
#' @return
#' * `summarize_coxreg()` returns a layout object suitable for passing to further layouting functions,
#'   or to [rtables::build_table()]. Adding this function to an `rtable` layout will add a Cox regression table
#'   containing the chosen statistics to the table layout.
#'
#' @seealso [fit_coxreg_univar()] and [fit_coxreg_multivar()] which also take the `variables`, `data`,
#'   `at` (univariate only), and `control` arguments but return unformatted univariate and multivariate
#'   Cox regression models, respectively.
#'
#' @examples
#' # summarize_coxreg
#'
#' result_univar <- basic_table() %>%
#'   summarize_coxreg(variables = u1_variables) %>%
#'   build_table(dta_bladder)
#' result_univar
#'
#' result_univar_covs <- basic_table() %>%
#'   summarize_coxreg(
#'     variables = u2_variables,
#'   ) %>%
#'   build_table(dta_bladder)
#' result_univar_covs
#'
#' result_multivar <- basic_table() %>%
#'   summarize_coxreg(
#'     variables = m1_variables,
#'     multivar = TRUE,
#'   ) %>%
#'   build_table(dta_bladder)
#' result_multivar
#'
#' result_multivar_covs <- basic_table() %>%
#'   summarize_coxreg(
#'     variables = m2_variables,
#'     multivar = TRUE,
#'     varlabels = c("Covariate 1", "Covariate 2") # custom labels
#'   ) %>%
#'   build_table(dta_bladder)
#' result_multivar_covs
#'
#' @export
#' @order 2
summarize_coxreg <- function(lyt,
                             variables,
                             control = control_coxreg(),
                             at = list(),
                             multivar = FALSE,
                             common_var = "STUDYID",
                             .stats = c("n", "hr", "ci", "pval", "pval_inter"),
                             .formats = c(
                               n = "xx", hr = "xx.xx", ci = "(xx.xx, xx.xx)",
                               pval = "x.xxxx | (<0.0001)", pval_inter = "x.xxxx | (<0.0001)"
                             ),
                             varlabels = NULL,
                             .indent_mods = NULL,
                             na_str = "",
                             .section_div = NA_character_) {
  if (multivar && control$interaction) {
    warning(paste(
      "Interactions are not available for multivariate cox regression using summarize_coxreg.",
      "The model will be calculated without interaction effects."
    ))
  }
  if (control$interaction && !"arm" %in% names(variables)) {
    stop("To include interactions please specify 'arm' in variables.")
  }

  .stats <- if (!"arm" %in% names(variables) || multivar) { # only valid statistics
    intersect(c("hr", "ci", "pval"), .stats)
  } else if (control$interaction) {
    intersect(c("n", "hr", "ci", "pval", "pval_inter"), .stats)
  } else {
    intersect(c("n", "hr", "ci", "pval"), .stats)
  }
  stat_labels <- c(
    n = "n", hr = "Hazard Ratio", ci = paste0(control$conf_level * 100, "% CI"),
    pval = "p-value", pval_inter = "Interaction p-value"
  )
  stat_labels <- stat_labels[names(stat_labels) %in% .stats]
  .formats <- .formats[names(.formats) %in% .stats]
  env <- new.env() # create caching environment

  lyt <- lyt %>%
    split_cols_by_multivar(
      vars = rep(common_var, length(.stats)),
      varlabels = stat_labels,
      extra_args = list(
        .stats = .stats, .formats = .formats, .indent_mods = .indent_mods, na_str = rep(na_str, length(.stats)),
        cache_env = replicate(length(.stats), list(env))
      )
    )

  if ("arm" %in% names(variables)) { # treatment effect
    lyt <- lyt %>%
      split_rows_by(
        common_var,
        split_label = "Treatment:",
        label_pos = "visible",
        child_labels = "hidden",
        section_div = head(.section_div, 1)
      )
    if (!multivar) {
      lyt <- lyt %>%
        analyze_colvars(
          afun = a_coxreg,
          na_str = na_str,
          extra_args = list(
            variables = variables, control = control, multivar = multivar, eff = TRUE, var_main = multivar,
            labelstr = ""
          )
        )
    } else { # treatment level effects
      lyt <- lyt %>%
        summarize_row_groups(
          cfun = a_coxreg,
          na_str = na_str,
          extra_args = list(
            variables = variables, control = control, multivar = multivar, eff = TRUE, var_main = multivar
          )
        ) %>%
        analyze_colvars(
          afun = a_coxreg,
          na_str = na_str,
          extra_args = list(eff = TRUE, control = control, variables = variables, multivar = multivar, labelstr = "")
        )
    }
  }

  if ("covariates" %in% names(variables)) { # covariate main effects
    lyt <- lyt %>%
      split_rows_by_multivar(
        vars = variables$covariates,
        varlabels = varlabels,
        split_label = "Covariate:",
        nested = FALSE,
        child_labels = if (multivar || control$interaction || !"arm" %in% names(variables)) "default" else "hidden",
        section_div = tail(.section_div, 1)
      )
    if (multivar || control$interaction || !"arm" %in% names(variables)) {
      lyt <- lyt %>%
        summarize_row_groups(
          cfun = a_coxreg,
          na_str = na_str,
          extra_args = list(
            variables = variables, at = at, control = control, multivar = multivar,
            var_main = if (multivar) multivar else control$interaction
          )
        )
    } else {
      if (!is.null(varlabels)) names(varlabels) <- variables$covariates
      lyt <- lyt %>%
        analyze_colvars(
          afun = a_coxreg,
          na_str = na_str,
          extra_args = list(
            variables = variables, at = at, control = control, multivar = multivar,
            var_main = if (multivar) multivar else control$interaction,
            labelstr = if (is.null(varlabels)) "" else varlabels
          )
        )
    }

    if (!"arm" %in% names(variables)) control$interaction <- TRUE # special case: univar no arm
    if (multivar || control$interaction) { # covariate level effects
      lyt <- lyt %>%
        analyze_colvars(
          afun = a_coxreg,
          na_str = na_str,
          extra_args = list(variables = variables, at = at, control = control, multivar = multivar, labelstr = ""),
          indent_mod = if (!"arm" %in% names(variables) || multivar) 0L else -1L
        )
    }
  }

  lyt
}

#' Count patients with abnormal analysis range values by baseline status
#'
#' @description `r lifecycle::badge("stable")`
#'
#' The analyze function [count_abnormal_by_baseline()] creates a layout element to count patients with abnormal
#' analysis range values, categorized by baseline status.
#'
#' This function analyzes primary analysis variable `var` which indicates abnormal range results. Additional
#' analysis variables that can be supplied as a list via the `variables` parameter are `id` (defaults to
#' `USUBJID`), a variable to indicate unique subject identifiers, and `baseline` (defaults to `BNRIND`), a
#' variable to indicate baseline reference ranges.
#'
#' For each direction specified via the `abnormal` parameter (e.g. High or Low), we condition on baseline
#' range result and count patients in the numerator and denominator as follows for each of the following
#' categories:
#'   * `Not <abnormality>`
#'     * `num`:  The number of patients without abnormality at baseline (excluding those with missing baseline)
#'       and with at least one abnormality post-baseline.
#'     * `denom`: The number of patients without abnormality at baseline (excluding those with missing baseline).
#'   * `<Abnormality>`
#'     * `num`: The number of patients with abnormality as baseline and at least one abnormality post-baseline.
#'     * `denom`: The number of patients with abnormality at baseline.
#'   * `Total`
#'     * `num`: The number of patients with at least one post-baseline record and at least one abnormality
#'       post-baseline.
#'     * `denom`: The number of patients with at least one post-baseline record.
#'
#' This function assumes that `df` has been filtered to only include post-baseline records.
#'
#' @inheritParams argument_convention
#' @param abnormal (`character`)\cr values identifying the abnormal range level(s) in `.var`.
#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("abnormal_by_baseline")`
#'   to see available statistics for this function.
#'
#' @note
#' * `df` should be filtered to include only post-baseline records.
#' * If the baseline variable or analysis variable contains `NA` records, it is expected that `df` has been
#'   pre-processed using [df_explicit_na()] or [explicit_na()].
#'
#' @seealso Relevant description function [d_count_abnormal_by_baseline()].
#'
#' @name abnormal_by_baseline
#' @order 1
NULL

#' Description function for `s_count_abnormal_by_baseline()`
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Description function that produces the labels for [s_count_abnormal_by_baseline()].
#'
#' @inheritParams abnormal_by_baseline
#'
#' @return Abnormal category labels for [s_count_abnormal_by_baseline()].
#'
#' @examples
#' d_count_abnormal_by_baseline("LOW")
#'
#' @export
d_count_abnormal_by_baseline <- function(abnormal) {
  not_abn_name <- paste("Not", tolower(abnormal))
  abn_name <- paste0(toupper(substr(abnormal, 1, 1)), tolower(substring(abnormal, 2)))
  total_name <- "Total"

  list(
    not_abnormal = not_abn_name,
    abnormal = abn_name,
    total = total_name
  )
}

#' @describeIn abnormal_by_baseline Statistics function for a single `abnormal` level.
#'
#' @param na_str (`string`)\cr the explicit `na_level` argument you used in the pre-processing steps (maybe with
#'   [df_explicit_na()]). The default is `"<Missing>"`.
#'
#' @return
#' * `s_count_abnormal_by_baseline()` returns statistic `fraction` which is a named list with 3 labeled elements:
#'   `not_abnormal`, `abnormal`, and `total`. Each element contains a vector with `num` and `denom` patient counts.
#'
#' @keywords internal
s_count_abnormal_by_baseline <- function(df,
                                         .var,
                                         abnormal,
                                         na_str = "<Missing>",
                                         variables = list(id = "USUBJID", baseline = "BNRIND")) {
  checkmate::assert_string(.var)
  checkmate::assert_string(abnormal)
  checkmate::assert_string(na_str)
  assert_df_with_variables(df, c(range = .var, variables))
  checkmate::assert_subset(names(variables), c("id", "baseline"))
  checkmate::assert_multi_class(df[[variables$id]], classes = c("factor", "character"))
  checkmate::assert_multi_class(df[[variables$baseline]], classes = c("factor", "character"))
  checkmate::assert_multi_class(df[[.var]], classes = c("factor", "character"))

  # If input is passed as character, changed to factor
  df[[.var]] <- as_factor_keep_attributes(df[[.var]], na_level = na_str)
  df[[variables$baseline]] <- as_factor_keep_attributes(df[[variables$baseline]], na_level = na_str)

  assert_valid_factor(df[[.var]], any.missing = FALSE)
  assert_valid_factor(df[[variables$baseline]], any.missing = FALSE)

  # Keep only records with valid analysis value.
  df <- df[df[[.var]] != na_str, ]

  anl <- data.frame(
    id = df[[variables$id]],
    var = df[[.var]],
    baseline = df[[variables$baseline]],
    stringsAsFactors = FALSE
  )

  # Total:
  #  - Patients in denominator: have at least one valid measurement post-baseline.
  #  - Patients in numerator: have at least one abnormality.
  total_denom <- length(unique(anl$id))
  total_num <- length(unique(anl$id[anl$var == abnormal]))

  # Baseline NA records are counted only in total rows.
  anl <- anl[anl$baseline != na_str, ]

  # Abnormal:
  #   - Patients in denominator: have abnormality at baseline.
  #   - Patients in numerator: have abnormality at baseline AND
  #     have at least one abnormality post-baseline.
  abn_denom <- length(unique(anl$id[anl$baseline == abnormal]))
  abn_num <- length(unique(anl$id[anl$baseline == abnormal & anl$var == abnormal]))

  # Not abnormal:
  #   - Patients in denominator: do not have abnormality at baseline.
  #   - Patients in numerator: do not have abnormality at baseline AND
  #     have at least one abnormality post-baseline.
  not_abn_denom <- length(unique(anl$id[anl$baseline != abnormal]))
  not_abn_num <- length(unique(anl$id[anl$baseline != abnormal & anl$var == abnormal]))

  labels <- d_count_abnormal_by_baseline(abnormal)
  list(fraction = list(
    not_abnormal = formatters::with_label(c(num = not_abn_num, denom = not_abn_denom), labels$not_abnormal),
    abnormal = formatters::with_label(c(num = abn_num, denom = abn_denom), labels$abnormal),
    total = formatters::with_label(c(num = total_num, denom = total_denom), labels$total)
  ))
}

#' @describeIn abnormal_by_baseline Formatted analysis function which is used as `afun`
#'   in `count_abnormal_by_baseline()`.
#'
#' @return
#' * `a_count_abnormal_by_baseline()` returns the corresponding list with formatted [rtables::CellValue()].
#'
#' @keywords internal
a_count_abnormal_by_baseline <- make_afun(
  s_count_abnormal_by_baseline,
  .formats = c(fraction = format_fraction)
)

#' @describeIn abnormal_by_baseline Layout-creating function which can take statistics function arguments
#'   and additional format arguments. This function is a wrapper for [rtables::analyze()].
#'
#' @return
#' * `count_abnormal_by_baseline()` returns a layout object suitable for passing to further layouting functions,
#'   or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
#'   the statistics from `s_count_abnormal_by_baseline()` to the table layout.
#'
#' @examples
#' df <- data.frame(
#'   USUBJID = as.character(c(1:6)),
#'   ANRIND = factor(c(rep("LOW", 4), "NORMAL", "HIGH")),
#'   BNRIND = factor(c("LOW", "NORMAL", "HIGH", NA, "LOW", "NORMAL"))
#' )
#' df <- df_explicit_na(df)
#'
#' # Layout creating function.
#' basic_table() %>%
#'   count_abnormal_by_baseline(var = "ANRIND", abnormal = c(High = "HIGH")) %>%
#'   build_table(df)
#'
#' # Passing of statistics function and formatting arguments.
#' df2 <- data.frame(
#'   ID = as.character(c(1, 2, 3, 4)),
#'   RANGE = factor(c("NORMAL", "LOW", "HIGH", "HIGH")),
#'   BLRANGE = factor(c("LOW", "HIGH", "HIGH", "NORMAL"))
#' )
#'
#' basic_table() %>%
#'   count_abnormal_by_baseline(
#'     var = "RANGE",
#'     abnormal = c(Low = "LOW"),
#'     variables = list(id = "ID", baseline = "BLRANGE"),
#'     .formats = c(fraction = "xx / xx"),
#'     .indent_mods = c(fraction = 2L)
#'   ) %>%
#'   build_table(df2)
#'
#' @export
#' @order 2
count_abnormal_by_baseline <- function(lyt,
                                       var,
                                       abnormal,
                                       variables = list(id = "USUBJID", baseline = "BNRIND"),
                                       na_str = "<Missing>",
                                       nested = TRUE,
                                       ...,
                                       table_names = abnormal,
                                       .stats = NULL,
                                       .formats = NULL,
                                       .labels = NULL,
                                       .indent_mods = NULL) {
  checkmate::assert_character(abnormal, len = length(table_names), names = "named")
  checkmate::assert_string(var)

  extra_args <- list(abnormal = abnormal, variables = variables, na_str = na_str, ...)

  afun <- make_afun(
    a_count_abnormal_by_baseline,
    .stats = .stats,
    .formats = .formats,
    .labels = .labels,
    .indent_mods = .indent_mods,
    .ungroup_stats = "fraction"
  )
  for (i in seq_along(abnormal)) {
    extra_args[["abnormal"]] <- abnormal[i]

    lyt <- analyze(
      lyt = lyt,
      vars = var,
      var_labels = names(abnormal[i]),
      afun = afun,
      na_str = na_str,
      nested = nested,
      table_names = table_names[i],
      extra_args = extra_args,
      show_labels = "visible"
    )
  }
  lyt
}

#' Add titles, footnotes, page Number, and a bounding box to a grid grob
#'
#' @description `r lifecycle::badge("stable")`
#'
#' This function is useful to label grid grobs (also `ggplot2`, and `lattice` plots)
#' with title, footnote, and page numbers.
#'
#' @inheritParams grid::grob
#' @param grob (`grob`)\cr a grid grob object, optionally `NULL` if only a `grob` with the decoration should be shown.
#' @param titles (`character`)\cr titles given as a vector of strings that are each separated by a newline and wrapped
#'   according to the page width.
#' @param footnotes (`character`)\cr footnotes. Uses the same formatting rules as `titles`.
#' @param page (`string` or `NULL`)\cr page numeration. If `NULL` then no page number is displayed.
#' @param width_titles (`grid::unit`)\cr width of titles. Usually defined as all the available space
#'   `grid::unit(1, "npc")`, it is affected by the parameter `outer_margins`. Right margins (`outer_margins[4]`)
#'   need to be subtracted to the allowed width.
#' @param width_footnotes (`grid::unit`)\cr width of footnotes. Same default and margin correction as `width_titles`.
#' @param border (`flag`)\cr whether a border should be drawn around the plot or not.
#' @param padding (`grid::unit`)\cr padding. A unit object of length 4. Innermost margin between the plot (`grob`)
#'   and, possibly, the border of the plot. Usually expressed in 4 identical values (usually `"lines"`). It defaults
#'   to `grid::unit(rep(1, 4), "lines")`.
#' @param margins (`grid::unit`)\cr margins. A unit object of length 4. Margins between the plot and the other
#'   elements in the list (e.g. titles, plot, and footers). This is usually expressed in 4 `"lines"`, where the
#'   lateral ones are 0s, while top and bottom are 1s. It defaults to `grid::unit(c(1, 0, 1, 0), "lines")`.
#' @param outer_margins (`grid::unit`)\cr outer margins. A unit object of length 4. It defines the general margin of
#'   the plot, considering also decorations like titles, footnotes, and page numbers. It defaults to
#'   `grid::unit(c(2, 1.5, 3, 1.5), "cm")`.
#' @param gp_titles (`gpar`)\cr a `gpar` object. Mainly used to set different `"fontsize"`.
#' @param gp_footnotes (`gpar`)\cr a `gpar` object. Mainly used to set different `"fontsize"`.
#'
#' @return A grid grob (`gTree`).
#'
#' @details The titles and footnotes will be ragged, i.e. each title will be wrapped individually.
#'
#' @examples
#' library(grid)
#'
#' titles <- c(
#'   "Edgar Anderson's Iris Data",
#'   paste(
#'     "This famous (Fisher's or Anderson's) iris data set gives the measurements",
#'     "in centimeters of the variables sepal length and width and petal length",
#'     "and width, respectively, for 50 flowers from each of 3 species of iris."
#'   )
#' )
#'
#' footnotes <- c(
#'   "The species are Iris setosa, versicolor, and virginica.",
#'   paste(
#'     "iris is a data frame with 150 cases (rows) and 5 variables (columns) named",
#'     "Sepal.Length, Sepal.Width, Petal.Length, Petal.Width, and Species."
#'   )
#' )
#'
#' ## empty plot
#' grid.newpage()
#'
#' grid.draw(
#'   decorate_grob(
#'     NULL,
#'     titles = titles,
#'     footnotes = footnotes,
#'     page = "Page 4 of 10"
#'   )
#' )
#'
#' # grid
#' p <- gTree(
#'   children = gList(
#'     rectGrob(),
#'     xaxisGrob(),
#'     yaxisGrob(),
#'     textGrob("Sepal.Length", y = unit(-4, "lines")),
#'     textGrob("Petal.Length", x = unit(-3.5, "lines"), rot = 90),
#'     pointsGrob(iris$Sepal.Length, iris$Petal.Length, gp = gpar(col = iris$Species), pch = 16)
#'   ),
#'   vp = vpStack(plotViewport(), dataViewport(xData = iris$Sepal.Length, yData = iris$Petal.Length))
#' )
#' grid.newpage()
#' grid.draw(p)
#'
#' grid.newpage()
#' grid.draw(
#'   decorate_grob(
#'     grob = p,
#'     titles = titles,
#'     footnotes = footnotes,
#'     page = "Page 6 of 129"
#'   )
#' )
#'
#' ## with ggplot2
#' library(ggplot2)
#'
#' p_gg <- ggplot2::ggplot(iris, aes(Sepal.Length, Sepal.Width, col = Species)) +
#'   ggplot2::geom_point()
#' p_gg
#' p <- ggplotGrob(p_gg)
#' grid.newpage()
#' grid.draw(
#'   decorate_grob(
#'     grob = p,
#'     titles = titles,
#'     footnotes = footnotes,
#'     page = "Page 6 of 129"
#'   )
#' )
#'
#' ## with lattice
#' library(lattice)
#'
#' xyplot(Sepal.Length ~ Petal.Length, data = iris, col = iris$Species)
#' p <- grid.grab()
#' grid.newpage()
#' grid.draw(
#'   decorate_grob(
#'     grob = p,
#'     titles = titles,
#'     footnotes = footnotes,
#'     page = "Page 6 of 129"
#'   )
#' )
#'
#' # with gridExtra - no borders
#' library(gridExtra)
#' grid.newpage()
#' grid.draw(
#'   decorate_grob(
#'     tableGrob(
#'       head(mtcars)
#'     ),
#'     titles = "title",
#'     footnotes = "footnote",
#'     border = FALSE
#'   )
#' )
#'
#' @export
decorate_grob <- function(grob,
                          titles,
                          footnotes,
                          page = "",
                          width_titles = grid::unit(1, "npc"),
                          width_footnotes = grid::unit(1, "npc"),
                          border = TRUE,
                          padding = grid::unit(rep(1, 4), "lines"),
                          margins = grid::unit(c(1, 0, 1, 0), "lines"),
                          outer_margins = grid::unit(c(2, 1.5, 3, 1.5), "cm"),
                          gp_titles = grid::gpar(),
                          gp_footnotes = grid::gpar(fontsize = 8),
                          name = NULL,
                          gp = grid::gpar(),
                          vp = NULL) {
  # External margins need to be taken into account when defining the width of titles and footers
  # because the text is split in advance depending on only the width of the viewport.
  if (any(as.numeric(outer_margins) > 0)) {
    width_titles <- width_titles - outer_margins[4]
    width_footnotes <- width_footnotes - outer_margins[4]
  }

  st_titles <- split_text_grob(
    titles,
    x = 0, y = 1,
    just = c("left", "top"),
    width = width_titles,
    vp = grid::viewport(layout.pos.row = 1, layout.pos.col = 1),
    gp = gp_titles
  )

  st_footnotes <- split_text_grob(
    footnotes,
    x = 0, y = 1,
    just = c("left", "top"),
    width = width_footnotes,
    vp = grid::viewport(layout.pos.row = 3, layout.pos.col = 1),
    gp = gp_footnotes
  )

  pg_footnote <- grid::textGrob(
    paste("\n", page),
    x = 1, y = 0,
    just = c("right", "bottom"),
    vp = grid::viewport(layout.pos.row = 4, layout.pos.col = 1),
    gp = gp_footnotes
  )

  # Initial decoration of the grob -> border, paddings, and margins are used here
  main_plot <- grid::gTree(
    children = grid::gList(
      if (border) grid::rectGrob(),
      grid::gTree(
        children = grid::gList(
          grob
        ),
        vp = grid::plotViewport(margins = padding) # innermost margins of the grob plot
      )
    ),
    vp = grid::vpStack(
      grid::viewport(layout.pos.row = 2, layout.pos.col = 1),
      grid::plotViewport(margins = margins) # margins around the border plot
    )
  )

  grid::gTree(
    grob = grob,
    titles = titles,
    footnotes = footnotes,
    page = page,
    width_titles = width_titles,
    width_footnotes = width_footnotes,
    outer_margins = outer_margins,
    gp_titles = gp_titles,
    gp_footnotes = gp_footnotes,
    children = grid::gList(
      grid::gTree(
        children = grid::gList(
          st_titles,
          main_plot, # main plot with border, padding, and margins
          st_footnotes,
          pg_footnote
        ),
        childrenvp = NULL,
        name = "titles_grob_footnotes",
        vp = grid::vpStack(
          grid::plotViewport(margins = outer_margins), # Main external margins
          grid::viewport(
            layout = grid::grid.layout(
              nrow = 4, ncol = 1,
              heights = grid::unit.c(
                grid::grobHeight(st_titles),
                grid::unit(1, "null"),
                grid::grobHeight(st_footnotes),
                grid::grobHeight(pg_footnote)
              )
            )
          )
        )
      )
    ),
    name = name,
    gp = gp,
    vp = vp,
    cl = "decoratedGrob"
  )
}

# nocov start
#' @importFrom grid validDetails
#' @noRd
validDetails.decoratedGrob <- function(x) {
  checkmate::assert_character(x$titles)
  checkmate::assert_character(x$footnotes)

  if (!is.null(x$grob)) {
    checkmate::assert_true(grid::is.grob(x$grob))
  }
  if (length(x$page) == 1) {
    checkmate::assert_character(x$page)
  }
  if (!grid::is.unit(x$outer_margins)) {
    checkmate::assert_vector(x$outer_margins, len = 4)
  }
  if (!grid::is.unit(x$margins)) {
    checkmate::assert_vector(x$margins, len = 4)
  }
  if (!grid::is.unit(x$padding)) {
    checkmate::assert_vector(x$padding, len = 4)
  }

  x
}

#' @importFrom grid widthDetails
#' @noRd
widthDetails.decoratedGrob <- function(x) {
  grid::unit(1, "null")
}

#' @importFrom grid heightDetails
#' @noRd
heightDetails.decoratedGrob <- function(x) {
  grid::unit(1, "null")
}

#' Split text according to available text width
#'
#' Dynamically wrap text.
#'
#' @inheritParams grid::grid.text
#' @param text (`string`)\cr the text to wrap.
#' @param width (`grid::unit`)\cr a unit object specifying maximum width of text.
#'
#' @return A text `grob`.
#'
#' @details This code is taken from `R Graphics by Paul Murell, 2nd edition`
#'
#' @keywords internal
split_text_grob <- function(text,
                            x = grid::unit(0.5, "npc"),
                            y = grid::unit(0.5, "npc"),
                            width = grid::unit(1, "npc"),
                            just = "centre",
                            hjust = NULL,
                            vjust = NULL,
                            default.units = "npc", # nolint
                            name = NULL,
                            gp = grid::gpar(),
                            vp = NULL) {
  text <- gsub("\\\\n", "\n", text) # fixing cases of mixed behavior (\n and \\n)

  if (!grid::is.unit(x)) x <- grid::unit(x, default.units)
  if (!grid::is.unit(y)) y <- grid::unit(y, default.units)
  if (!grid::is.unit(width)) width <- grid::unit(width, default.units)
  if (grid::unitType(x) %in% c("sum", "min", "max")) x <- grid::convertUnit(x, default.units)
  if (grid::unitType(y) %in% c("sum", "min", "max")) y <- grid::convertUnit(y, default.units)
  if (grid::unitType(width) %in% c("sum", "min", "max")) width <- grid::convertUnit(width, default.units)

  if (length(gp) > 0) { # account for effect of gp on text width -> it was bugging when text was empty
    horizontal_npc_width_no_gp <- grid::convertWidth(
      grid::grobWidth(
        grid::textGrob(
          paste0(text, collapse = "\n")
        )
      ), "npc",
      valueOnly = TRUE
    )
    horizontal_npc_width_with_gp <- grid::convertWidth(grid::grobWidth(
      grid::textGrob(
        paste0(text, collapse = "\n"),
        gp = gp
      )
    ), "npc", valueOnly = TRUE)

    # Adapting width to the input gpar (it is normalized so does not matter what is text)
    width <- width * horizontal_npc_width_no_gp / horizontal_npc_width_with_gp
  }

  ## if it is a fixed unit then we do not need to recalculate when viewport resized
  if (!inherits(width, "unit.arithmetic") && !is.null(attr(width, "unit")) &&
    attr(width, "unit") %in% c("cm", "inches", "mm", "points", "picas", "bigpts", "dida", "cicero", "scaledpts")) { # nolint
    attr(text, "fixed_text") <- paste(vapply(text, split_string, character(1), width = width), collapse = "\n")
  }

  # Fix for split_string in case of residual \n (otherwise is counted as character)
  text2 <- unlist(
    strsplit(
      paste0(text, collapse = "\n"), # for "" cases
      "\n"
    )
  )

  # Final grid text with cat-friendly split_string
  grid::grid.text(
    label = split_string(text2, width),
    x = x, y = y,
    just = just,
    hjust = hjust,
    vjust = vjust,
    rot = 0,
    check.overlap = FALSE,
    name = name,
    gp = gp,
    vp = vp,
    draw = FALSE
  )
}

#' @importFrom grid validDetails
#' @noRd
validDetails.dynamicSplitText <- function(x) {
  checkmate::assert_character(x$text)
  checkmate::assert_true(grid::is.unit(x$width))
  checkmate::assert_vector(x$width, len = 1)
  x
}

#' @importFrom grid heightDetails
#' @noRd
heightDetails.dynamicSplitText <- function(x) {
  txt <- if (!is.null(attr(x$text, "fixed_text"))) {
    attr(x$text, "fixed_text")
  } else {
    paste(vapply(x$text, split_string, character(1), width = x$width), collapse = "\n")
  }
  grid::stringHeight(txt)
}

#' @importFrom grid widthDetails
#' @noRd
widthDetails.dynamicSplitText <- function(x) {
  x$width
}

#' @importFrom grid drawDetails
#' @noRd
drawDetails.dynamicSplitText <- function(x, recording) {
  txt <- if (!is.null(attr(x$text, "fixed_text"))) {
    attr(x$text, "fixed_text")
  } else {
    paste(vapply(x$text, split_string, character(1), width = x$width), collapse = "\n")
  }

  x$width <- NULL
  x$label <- txt
  x$text <- NULL
  class(x) <- c("text", class(x)[-1])

  grid::grid.draw(x)
}
# nocov end

# Adapted from Paul Murell R Graphics 2nd Edition
# https://www.stat.auckland.ac.nz/~paul/RG2e/interactgrid-splittext.R
split_string <- function(text, width) {
  strings <- strsplit(text, " ")
  out_string <- NA
  for (string_i in seq_along(strings)) {
    newline_str <- strings[[string_i]]
    if (length(newline_str) == 0) newline_str <- ""
    if (is.na(out_string[string_i])) {
      out_string[string_i] <- newline_str[[1]][[1]]
      linewidth <- grid::stringWidth(out_string[string_i])
    }
    gapwidth <- grid::stringWidth(" ")
    availwidth <- as.numeric(width)
    if (length(newline_str) > 1) {
      for (i in seq(2, length(newline_str))) {
        width_i <- grid::stringWidth(newline_str[i])
        # Main conversion of allowed text width -> npc units are 0<npc<1. External viewport is used for conversion
        if (grid::convertWidth(linewidth + gapwidth + width_i, grid::unitType(width), valueOnly = TRUE) < availwidth) {
          sep <- " "
          linewidth <- linewidth + gapwidth + width_i
        } else {
          sep <- "\n"
          linewidth <- width_i
        }
        out_string[string_i] <- paste(out_string[string_i], newline_str[i], sep = sep)
      }
    }
  }
  paste(out_string, collapse = "\n")
}

#' Update page number
#'
#' Automatically updates page number.
#'
#' @param npages (`numeric(1)`)\cr total number of pages.
#' @param ... arguments passed on to [decorate_grob()].
#'
#' @return Closure that increments the page number.
#'
#' @keywords internal
decorate_grob_factory <- function(npages, ...) {
  current_page <- 0
  function(grob) {
    current_page <<- current_page + 1
    if (current_page > npages) {
      stop(paste("current page is", current_page, "but max.", npages, "specified."))
    }
    decorate_grob(grob = grob, page = paste("Page", current_page, "of", npages), ...)
  }
}

#' Decorate set of `grob`s and add page numbering
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Note that this uses the [decorate_grob_factory()] function.
#'
#' @param grobs (`list` of `grob`)\cr a list of grid grobs.
#' @param ... arguments passed on to [decorate_grob()].
#'
#' @return A decorated grob.
#'
#' @examples
#' library(ggplot2)
#' library(grid)
#' g <- with(data = iris, {
#'   list(
#'     ggplot2::ggplotGrob(
#'       ggplot2::ggplot(mapping = aes(Sepal.Length, Sepal.Width, col = Species)) +
#'         ggplot2::geom_point()
#'     ),
#'     ggplot2::ggplotGrob(
#'       ggplot2::ggplot(mapping = aes(Sepal.Length, Petal.Length, col = Species)) +
#'         ggplot2::geom_point()
#'     ),
#'     ggplot2::ggplotGrob(
#'       ggplot2::ggplot(mapping = aes(Sepal.Length, Petal.Width, col = Species)) +
#'         ggplot2::geom_point()
#'     ),
#'     ggplot2::ggplotGrob(
#'       ggplot2::ggplot(mapping = aes(Sepal.Width, Petal.Length, col = Species)) +
#'         ggplot2::geom_point()
#'     ),
#'     ggplot2::ggplotGrob(
#'       ggplot2::ggplot(mapping = aes(Sepal.Width, Petal.Width, col = Species)) +
#'         ggplot2::geom_point()
#'     ),
#'     ggplot2::ggplotGrob(
#'       ggplot2::ggplot(mapping = aes(Petal.Length, Petal.Width, col = Species)) +
#'         ggplot2::geom_point()
#'     )
#'   )
#' })
#' lg <- decorate_grob_set(grobs = g, titles = "Hello\nOne\nTwo\nThree", footnotes = "")
#'
#' draw_grob(lg[[1]])
#' draw_grob(lg[[2]])
#' draw_grob(lg[[6]])
#'
#' @export
decorate_grob_set <- function(grobs, ...) {
  n <- length(grobs)
  lgf <- decorate_grob_factory(npages = n, ...)
  lapply(grobs, lgf)
}

#' Survival time point analysis
#'
#' @description `r lifecycle::badge("stable")`
#'
#' The analyze function [surv_timepoint()] creates a layout element to analyze patient survival rates and difference
#' of survival rates between groups at a given time point. The primary analysis variable `vars` is the time variable.
#' Other required inputs are `time_point`, the numeric time point of interest, and `is_event`, a variable that
#' indicates whether or not an event has occurred. The `method` argument is used to specify whether you want to analyze
#' survival estimations (`"surv"`), difference in survival with the control (`"surv_diff"`), or both of these
#' (`"both"`).
#'
#' @inheritParams argument_convention
#' @inheritParams s_surv_time
#' @param time_point (`numeric(1)`)\cr survival time point of interest.
#' @param control (`list`)\cr parameters for comparison details, specified by using the helper function
#'   [control_surv_timepoint()]. Some possible parameter options are:
#'   * `conf_level` (`proportion`)\cr confidence level of the interval for survival rate.
#'   * `conf_type` (`string`)\cr confidence interval type. Options are "plain" (default), "log", "log-log",
#'     see more in [survival::survfit()]. Note option "none" is no longer supported.
#' @param method (`string`)\cr `"surv"` (survival estimations), `"surv_diff"` (difference in survival with the
#'   control), or `"both"`.
#' @param table_names_suffix (`string`)\cr optional suffix for the `table_names` used for the `rtables` to
#'   avoid warnings from duplicate table names.
#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("surv_timepoint")`
#'   to see available statistics for this function.
#' @param .indent_mods (named `integer`)\cr indent modifiers for the labels. Each element of the vector
#'   should be a name-value pair with name corresponding to a statistic specified in `.stats` and value the indentation
#'   for that statistic's row label.
#'
#' @name survival_timepoint
#' @order 1
NULL

#' @describeIn survival_timepoint Statistics function which analyzes survival rate.
#'
#' @return
#' * `s_surv_timepoint()` returns the statistics:
#'   * `pt_at_risk`: Patients remaining at risk.
#'   * `event_free_rate`: Event-free rate (%).
#'   * `rate_se`: Standard error of event free rate.
#'   * `rate_ci`: Confidence interval for event free rate.
#'
#' @keywords internal
s_surv_timepoint <- function(df,
                             .var,
                             time_point,
                             is_event,
                             control = control_surv_timepoint()) {
  checkmate::assert_string(.var)
  assert_df_with_variables(df, list(tte = .var, is_event = is_event))
  checkmate::assert_numeric(df[[.var]], min.len = 1, any.missing = FALSE)
  checkmate::assert_number(time_point)
  checkmate::assert_logical(df[[is_event]], min.len = 1, any.missing = FALSE)

  conf_type <- control$conf_type
  conf_level <- control$conf_level

  formula <- stats::as.formula(paste0("survival::Surv(", .var, ", ", is_event, ") ~ 1"))
  srv_fit <- survival::survfit(
    formula = formula,
    data = df,
    conf.int = conf_level,
    conf.type = conf_type
  )
  s_srv_fit <- summary(srv_fit, times = time_point, extend = TRUE)
  df_srv_fit <- as.data.frame(s_srv_fit[c("time", "n.risk", "surv", "lower", "upper", "std.err")])
  if (df_srv_fit[["n.risk"]] == 0) {
    pt_at_risk <- event_free_rate <- rate_se <- NA_real_
    rate_ci <- c(NA_real_, NA_real_)
  } else {
    pt_at_risk <- df_srv_fit$n.risk
    event_free_rate <- df_srv_fit$surv
    rate_se <- df_srv_fit$std.err
    rate_ci <- c(df_srv_fit$lower, df_srv_fit$upper)
  }
  list(
    pt_at_risk = formatters::with_label(pt_at_risk, "Patients remaining at risk"),
    event_free_rate = formatters::with_label(event_free_rate * 100, "Event Free Rate (%)"),
    rate_se = formatters::with_label(rate_se * 100, "Standard Error of Event Free Rate"),
    rate_ci = formatters::with_label(rate_ci * 100, f_conf_level(conf_level))
  )
}

#' @describeIn survival_timepoint Formatted analysis function which is used as `afun` in `surv_timepoint()`
#'   when `method = "surv"`.
#'
#' @return
#' * `a_surv_timepoint()` returns the corresponding list with formatted [rtables::CellValue()].
#'
#' @keywords internal
a_surv_timepoint <- make_afun(
  s_surv_timepoint,
  .indent_mods = c(
    pt_at_risk = 0L,
    event_free_rate = 0L,
    rate_se = 1L,
    rate_ci = 1L
  ),
  .formats = c(
    pt_at_risk = "xx",
    event_free_rate = "xx.xx",
    rate_se = "xx.xx",
    rate_ci = "(xx.xx, xx.xx)"
  )
)

#' @describeIn survival_timepoint Statistics function which analyzes difference between two survival rates.
#'
#' @return
#' * `s_surv_timepoint_diff()` returns the statistics:
#'   * `rate_diff`: Event-free rate difference between two groups.
#'   * `rate_diff_ci`: Confidence interval for the difference.
#'   * `ztest_pval`: p-value to test the difference is 0.
#'
#' @keywords internal
s_surv_timepoint_diff <- function(df,
                                  .var,
                                  .ref_group,
                                  .in_ref_col,
                                  time_point,
                                  control = control_surv_timepoint(),
                                  ...) {
  if (.in_ref_col) {
    return(
      list(
        rate_diff = formatters::with_label("", "Difference in Event Free Rate"),
        rate_diff_ci = formatters::with_label("", f_conf_level(control$conf_level)),
        ztest_pval = formatters::with_label("", "p-value (Z-test)")
      )
    )
  }
  data <- rbind(.ref_group, df)
  group <- factor(rep(c("ref", "x"), c(nrow(.ref_group), nrow(df))), levels = c("ref", "x"))
  res_per_group <- lapply(split(data, group), function(x) {
    s_surv_timepoint(df = x, .var = .var, time_point = time_point, control = control, ...)
  })

  res_x <- res_per_group[[2]]
  res_ref <- res_per_group[[1]]
  rate_diff <- res_x$event_free_rate - res_ref$event_free_rate
  se_diff <- sqrt(res_x$rate_se^2 + res_ref$rate_se^2)

  qs <- c(-1, 1) * stats::qnorm(1 - (1 - control$conf_level) / 2)
  rate_diff_ci <- rate_diff + qs * se_diff
  ztest_pval <- if (is.na(rate_diff)) {
    NA
  } else {
    2 * (1 - stats::pnorm(abs(rate_diff) / se_diff))
  }
  list(
    rate_diff = formatters::with_label(rate_diff, "Difference in Event Free Rate"),
    rate_diff_ci = formatters::with_label(rate_diff_ci, f_conf_level(control$conf_level)),
    ztest_pval = formatters::with_label(ztest_pval, "p-value (Z-test)")
  )
}

#' @describeIn survival_timepoint Formatted analysis function which is used as `afun` in `surv_timepoint()`
#'   when `method = "surv_diff"`.
#'
#' @return
#' * `a_surv_timepoint_diff()` returns the corresponding list with formatted [rtables::CellValue()].
#'
#' @keywords internal
a_surv_timepoint_diff <- make_afun(
  s_surv_timepoint_diff,
  .formats = c(
    rate_diff = "xx.xx",
    rate_diff_ci = "(xx.xx, xx.xx)",
    ztest_pval = "x.xxxx | (<0.0001)"
  )
)

#' @describeIn survival_timepoint Layout-creating function which can take statistics function arguments
#'   and additional format arguments. This function is a wrapper for [rtables::analyze()].
#'
#' @return
#' * `surv_timepoint()` returns a layout object suitable for passing to further layouting functions,
#'   or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
#'   the statistics from `s_surv_timepoint()` and/or `s_surv_timepoint_diff()` to the table layout depending on
#'   the value of `method`.
#'
#' @examples
#' library(dplyr)
#'
#' adtte_f <- tern_ex_adtte %>%
#'   filter(PARAMCD == "OS") %>%
#'   mutate(
#'     AVAL = day2month(AVAL),
#'     is_event = CNSR == 0
#'   )
#'
#' # Survival at given time points.
#' basic_table() %>%
#'   split_cols_by(var = "ARMCD", ref_group = "ARM A") %>%
#'   add_colcounts() %>%
#'   surv_timepoint(
#'     vars = "AVAL",
#'     var_labels = "Months",
#'     is_event = "is_event",
#'     time_point = 7
#'   ) %>%
#'   build_table(df = adtte_f)
#'
#' # Difference in survival at given time points.
#' basic_table() %>%
#'   split_cols_by(var = "ARMCD", ref_group = "ARM A") %>%
#'   add_colcounts() %>%
#'   surv_timepoint(
#'     vars = "AVAL",
#'     var_labels = "Months",
#'     is_event = "is_event",
#'     time_point = 9,
#'     method = "surv_diff",
#'     .indent_mods = c("rate_diff" = 0L, "rate_diff_ci" = 2L, "ztest_pval" = 2L)
#'   ) %>%
#'   build_table(df = adtte_f)
#'
#' # Survival and difference in survival at given time points.
#' basic_table() %>%
#'   split_cols_by(var = "ARMCD", ref_group = "ARM A") %>%
#'   add_colcounts() %>%
#'   surv_timepoint(
#'     vars = "AVAL",
#'     var_labels = "Months",
#'     is_event = "is_event",
#'     time_point = 9,
#'     method = "both"
#'   ) %>%
#'   build_table(df = adtte_f)
#'
#' @export
#' @order 2
surv_timepoint <- function(lyt,
                           vars,
                           time_point,
                           is_event,
                           control = control_surv_timepoint(),
                           method = c("surv", "surv_diff", "both"),
                           na_str = default_na_str(),
                           nested = TRUE,
                           ...,
                           table_names_suffix = "",
                           var_labels = "Time",
                           show_labels = "visible",
                           .stats = c(
                             "pt_at_risk", "event_free_rate", "rate_ci",
                             "rate_diff", "rate_diff_ci", "ztest_pval"
                           ),
                           .formats = NULL,
                           .labels = NULL,
                           .indent_mods = if (method == "both") {
                             c(rate_diff = 1L, rate_diff_ci = 2L, ztest_pval = 2L)
                           } else {
                             c(rate_diff_ci = 1L, ztest_pval = 1L)
                           }) {
  method <- match.arg(method)
  checkmate::assert_string(table_names_suffix)

  extra_args <- list(time_point = time_point, is_event = is_event, control = control, ...)

  f <- list(
    surv = c("pt_at_risk", "event_free_rate", "rate_se", "rate_ci"),
    surv_diff = c("rate_diff", "rate_diff_ci", "ztest_pval")
  )
  .stats <- h_split_param(.stats, .stats, f = f)
  .formats <- h_split_param(.formats, names(.formats), f = f)
  .labels <- h_split_param(.labels, names(.labels), f = f)
  .indent_mods <- h_split_param(.indent_mods, names(.indent_mods), f = f)

  afun_surv <- make_afun(
    a_surv_timepoint,
    .stats = .stats$surv,
    .formats = .formats$surv,
    .labels = .labels$surv,
    .indent_mods = .indent_mods$surv
  )

  afun_surv_diff <- make_afun(
    a_surv_timepoint_diff,
    .stats = .stats$surv_diff,
    .formats = .formats$surv_diff,
    .labels = .labels$surv_diff,
    .indent_mods = .indent_mods$surv_diff
  )

  time_point <- extra_args$time_point

  for (i in seq_along(time_point)) {
    extra_args[["time_point"]] <- time_point[i]

    if (method %in% c("surv", "both")) {
      lyt <- analyze(
        lyt,
        vars,
        var_labels = paste(time_point[i], var_labels),
        table_names = paste0("surv_", time_point[i], table_names_suffix),
        show_labels = show_labels,
        afun = afun_surv,
        na_str = na_str,
        nested = nested,
        extra_args = extra_args
      )
    }

    if (method %in% c("surv_diff", "both")) {
      lyt <- analyze(
        lyt,
        vars,
        var_labels = paste(time_point[i], var_labels),
        table_names = paste0("surv_diff_", time_point[i], table_names_suffix),
        show_labels = ifelse(method == "both", "hidden", show_labels),
        afun = afun_surv_diff,
        na_str = na_str,
        nested = nested,
        extra_args = extra_args
      )
    }
  }
  lyt
}

#' Create a STEP graph
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Based on the STEP results, creates a `ggplot` graph showing the estimated HR or OR
#' along the continuous biomarker value subgroups.
#'
#' @param df (`tibble`)\cr result of [tidy.step()].
#' @param use_percentile (`flag`)\cr whether to use percentiles for the x axis or actual
#'   biomarker values.
#' @param est (named `list`)\cr `col` and `lty` settings for estimate line.
#' @param ci_ribbon (named `list` or `NULL`)\cr `fill` and `alpha` settings for the confidence interval
#'   ribbon area, or `NULL` to not plot a CI ribbon.
#' @param col (`character`)\cr color(s).
#'
#' @return A `ggplot` STEP graph.
#'
#' @seealso Custom tidy method [tidy.step()].
#'
#' @examples
#' library(nestcolor)
#' library(survival)
#' lung$sex <- factor(lung$sex)
#'
#' # Survival example.
#' vars <- list(
#'   time = "time",
#'   event = "status",
#'   arm = "sex",
#'   biomarker = "age"
#' )
#'
#' step_matrix <- fit_survival_step(
#'   variables = vars,
#'   data = lung,
#'   control = c(control_coxph(), control_step(num_points = 10, degree = 2))
#' )
#' step_data <- broom::tidy(step_matrix)
#'
#' # Default plot.
#' g_step(step_data)
#'
#' # Add the reference 1 horizontal line.
#' library(ggplot2)
#' g_step(step_data) +
#'   ggplot2::geom_hline(ggplot2::aes(yintercept = 1), linetype = 2)
#'
#' # Use actual values instead of percentiles, different color for estimate and no CI,
#' # use log scale for y axis.
#' g_step(
#'   step_data,
#'   use_percentile = FALSE,
#'   est = list(col = "blue", lty = 1),
#'   ci_ribbon = NULL
#' ) + scale_y_log10()
#'
#' # Adding another curve based on additional column.
#' step_data$extra <- exp(step_data$`Percentile Center`)
#' g_step(step_data) +
#'   ggplot2::geom_line(ggplot2::aes(y = extra), linetype = 2, color = "green")
#'
#' # Response example.
#' vars <- list(
#'   response = "status",
#'   arm = "sex",
#'   biomarker = "age"
#' )
#'
#' step_matrix <- fit_rsp_step(
#'   variables = vars,
#'   data = lung,
#'   control = c(
#'     control_logistic(response_definition = "I(response == 2)"),
#'     control_step()
#'   )
#' )
#' step_data <- broom::tidy(step_matrix)
#' g_step(step_data)
#'
#' @export
g_step <- function(df,
                   use_percentile = "Percentile Center" %in% names(df),
                   est = list(col = "blue", lty = 1),
                   ci_ribbon = list(fill = getOption("ggplot2.discrete.colour")[1], alpha = 0.5),
                   col = getOption("ggplot2.discrete.colour")) {
  checkmate::assert_tibble(df)
  checkmate::assert_flag(use_percentile)
  checkmate::assert_character(col, null.ok = TRUE)
  checkmate::assert_list(est, names = "named")
  checkmate::assert_list(ci_ribbon, names = "named", null.ok = TRUE)

  x_var <- ifelse(use_percentile, "Percentile Center", "Interval Center")
  df$x <- df[[x_var]]
  attrs <- attributes(df)
  df$y <- df[[attrs$estimate]]

  # Set legend names. To be modified also at call level
  legend_names <- c("Estimate", "CI 95%")

  p <- ggplot2::ggplot(df, ggplot2::aes(x = .data[["x"]], y = .data[["y"]]))

  if (!is.null(col)) {
    p <- p +
      ggplot2::scale_color_manual(values = col)
  }

  if (!is.null(ci_ribbon)) {
    if (is.null(ci_ribbon$fill)) {
      ci_ribbon$fill <- "lightblue"
    }
    p <- p + ggplot2::geom_ribbon(
      ggplot2::aes(
        ymin = .data[["ci_lower"]], ymax = .data[["ci_upper"]],
        fill = legend_names[2]
      ),
      alpha = ci_ribbon$alpha
    ) +
      scale_fill_manual(
        name = "", values = c("CI 95%" = ci_ribbon$fill)
      )
  }
  suppressMessages(p <- p +
    ggplot2::geom_line(
      ggplot2::aes(y = .data[["y"]], color = legend_names[1]),
      linetype = est$lty
    ) +
    scale_colour_manual(
      name = "", values = c("Estimate" = "blue")
    ))

  p <- p + ggplot2::labs(x = attrs$biomarker, y = attrs$estimate)
  if (use_percentile) {
    p <- p + ggplot2::scale_x_continuous(labels = scales::percent)
  }
  p
}

#' Custom tidy method for STEP results
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Tidy the STEP results into a `tibble` format ready for plotting.
#'
#' @param x (`matrix`)\cr results from [fit_survival_step()].
#' @param ... not used.
#'
#' @return A `tibble` with one row per STEP subgroup. The estimates and CIs are on the HR or OR scale,
#'   respectively. Additional attributes carry metadata also used for plotting.
#'
#' @seealso [g_step()] which consumes the result from this function.
#'
#' @method tidy step
#'
#' @examples
#' library(survival)
#' lung$sex <- factor(lung$sex)
#' vars <- list(
#'   time = "time",
#'   event = "status",
#'   arm = "sex",
#'   biomarker = "age"
#' )
#' step_matrix <- fit_survival_step(
#'   variables = vars,
#'   data = lung,
#'   control = c(control_coxph(), control_step(num_points = 10, degree = 2))
#' )
#' broom::tidy(step_matrix)
#'
#' @export
tidy.step <- function(x, ...) { # nolint
  checkmate::assert_class(x, "step")
  dat <- as.data.frame(x)
  nams <- names(dat)
  is_surv <- "loghr" %in% names(dat)
  est_var <- ifelse(is_surv, "loghr", "logor")
  new_est_var <- ifelse(is_surv, "Hazard Ratio", "Odds Ratio")
  new_y_vars <- c(new_est_var, c("ci_lower", "ci_upper"))
  names(dat)[match(est_var, nams)] <- new_est_var
  dat[, new_y_vars] <- exp(dat[, new_y_vars])
  any_is_na <- any(is.na(dat[, new_y_vars]))
  any_is_very_large <- any(abs(dat[, new_y_vars]) > 1e10, na.rm = TRUE)
  if (any_is_na) {
    warning(paste(
      "Missing values in the point estimate or CI columns,",
      "this will lead to holes in the `g_step()` plot"
    ))
  }
  if (any_is_very_large) {
    warning(paste(
      "Very large absolute values in the point estimate or CI columns,",
      "consider adding `scale_y_log10()` to the `g_step()` result for plotting"
    ))
  }
  if (any_is_na || any_is_very_large) {
    warning("Consider using larger `bandwidth`, less `num_points` in `control_step()` settings for fitting")
  }
  structure(
    tibble::as_tibble(dat),
    estimate = new_est_var,
    biomarker = attr(x, "variables")$biomarker,
    ci = f_conf_level(attr(x, "control")$conf_level)
  )
}

#' Control function for Cox regression
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Sets a list of parameters for Cox regression fit. Used internally.
#'
#' @inheritParams argument_convention
#' @param pval_method (`string`)\cr the method used for estimation of p.values; `wald` (default) or `likelihood`.
#' @param interaction (`flag`)\cr if `TRUE`, the model includes the interaction between the studied
#'   treatment and candidate covariate. Note that for univariate models without treatment arm, and
#'   multivariate models, no interaction can be used so that this needs to be `FALSE`.
#' @param ties (`string`)\cr among `exact` (equivalent to `DISCRETE` in SAS), `efron` and `breslow`,
#'   see [survival::coxph()]. Note: there is no equivalent of SAS `EXACT` method in R.
#'
#' @return A `list` of items with names corresponding to the arguments.
#'
#' @seealso [fit_coxreg_univar()] and [fit_coxreg_multivar()].
#'
#' @examples
#' control_coxreg()
#'
#' @export
control_coxreg <- function(pval_method = c("wald", "likelihood"),
                           ties = c("exact", "efron", "breslow"),
                           conf_level = 0.95,
                           interaction = FALSE) {
  pval_method <- match.arg(pval_method)
  ties <- match.arg(ties)
  checkmate::assert_flag(interaction)
  assert_proportion_value(conf_level)
  list(
    pval_method = pval_method,
    ties = ties,
    conf_level = conf_level,
    interaction = interaction
  )
}

#' Custom tidy methods for Cox regression
#'
#' @description `r lifecycle::badge("stable")`
#'
#' @inheritParams argument_convention
#' @param x (`list`)\cr result of the Cox regression model fitted by [fit_coxreg_univar()] (for univariate models)
#'   or [fit_coxreg_multivar()] (for multivariate models).
#'
#' @return [broom::tidy()] returns:
#' * For `summary.coxph` objects,  a `data.frame` with columns: `Pr(>|z|)`, `exp(coef)`, `exp(-coef)`, `lower .95`,
#'   `upper .95`, `level`, and `n`.
#' * For `coxreg.univar` objects, a `data.frame` with columns: `effect`, `term`, `term_label`, `level`, `n`, `hr`,
#'   `lcl`, `ucl`, `pval`, and `ci`.
#' * For `coxreg.multivar` objects, a `data.frame` with columns: `term`, `pval`, `term_label`, `hr`, `lcl`, `ucl`,
#'   `level`, and `ci`.
#'
#' @seealso [cox_regression]
#'
#' @name tidy_coxreg
NULL

#' @describeIn tidy_coxreg Custom tidy method for [survival::coxph()] summary results.
#'
#' Tidy the [survival::coxph()] results into a `data.frame` to extract model results.
#'
#' @method tidy summary.coxph
#'
#' @examples
#' library(survival)
#' library(broom)
#'
#' set.seed(1, kind = "Mersenne-Twister")
#'
#' dta_bladder <- with(
#'   data = bladder[bladder$enum < 5, ],
#'   data.frame(
#'     time = stop,
#'     status = event,
#'     armcd = as.factor(rx),
#'     covar1 = as.factor(enum),
#'     covar2 = factor(
#'       sample(as.factor(enum)),
#'       levels = 1:4, labels = c("F", "F", "M", "M")
#'     )
#'   )
#' )
#' labels <- c("armcd" = "ARM", "covar1" = "A Covariate Label", "covar2" = "Sex (F/M)")
#' formatters::var_labels(dta_bladder)[names(labels)] <- labels
#' dta_bladder$age <- sample(20:60, size = nrow(dta_bladder), replace = TRUE)
#'
#' formula <- "survival::Surv(time, status) ~ armcd + covar1"
#' msum <- summary(coxph(stats::as.formula(formula), data = dta_bladder))
#' tidy(msum)
#'
#' @export
tidy.summary.coxph <- function(x, # nolint
                               ...) {
  checkmate::assert_class(x, "summary.coxph")
  pval <- x$coefficients
  confint <- x$conf.int
  levels <- rownames(pval)

  pval <- tibble::as_tibble(pval)
  confint <- tibble::as_tibble(confint)

  ret <- cbind(pval[, grepl("Pr", names(pval))], confint)
  ret$level <- levels
  ret$n <- x[["n"]]
  ret
}

#' @describeIn tidy_coxreg Custom tidy method for a univariate Cox regression.
#'
#' Tidy up the result of a Cox regression model fitted by [fit_coxreg_univar()].
#'
#' @method tidy coxreg.univar
#'
#' @examples
#' ## Cox regression: arm + 1 covariate.
#' mod1 <- fit_coxreg_univar(
#'   variables = list(
#'     time = "time", event = "status", arm = "armcd",
#'     covariates = "covar1"
#'   ),
#'   data = dta_bladder,
#'   control = control_coxreg(conf_level = 0.91)
#' )
#'
#' ## Cox regression: arm + 1 covariate + interaction, 2 candidate covariates.
#' mod2 <- fit_coxreg_univar(
#'   variables = list(
#'     time = "time", event = "status", arm = "armcd",
#'     covariates = c("covar1", "covar2")
#'   ),
#'   data = dta_bladder,
#'   control = control_coxreg(conf_level = 0.91, interaction = TRUE)
#' )
#'
#' tidy(mod1)
#' tidy(mod2)
#'
#' @export
tidy.coxreg.univar <- function(x, # nolint
                               ...) {
  checkmate::assert_class(x, "coxreg.univar")
  mod <- x$mod
  vars <- c(x$vars$arm, x$vars$covariates)
  has_arm <- "arm" %in% names(x$vars)

  result <- if (!has_arm) {
    Map(
      mod = mod, vars = vars,
      f = function(mod, vars) {
        h_coxreg_multivar_extract(
          var = vars,
          data = x$data,
          mod = mod,
          control = x$control
        )
      }
    )
  } else if (x$control$interaction) {
    Map(
      mod = mod, covar = vars,
      f = function(mod, covar) {
        h_coxreg_extract_interaction(
          effect = x$vars$arm, covar = covar, mod = mod, data = x$data,
          at = x$at, control = x$control
        )
      }
    )
  } else {
    Map(
      mod = mod, vars = vars,
      f = function(mod, vars) {
        h_coxreg_univar_extract(
          effect = x$vars$arm, covar = vars, data = x$data, mod = mod,
          control = x$control
        )
      }
    )
  }
  result <- do.call(rbind, result)

  result$ci <- Map(lcl = result$lcl, ucl = result$ucl, f = function(lcl, ucl) c(lcl, ucl))
  result$n <- lapply(result$n, empty_vector_if_na)
  result$ci <- lapply(result$ci, empty_vector_if_na)
  result$hr <- lapply(result$hr, empty_vector_if_na)
  if (x$control$interaction) {
    result$pval_inter <- lapply(result$pval_inter, empty_vector_if_na)
    # Remove interaction p-values due to change in specifications.
    result$pval[result$effect != "Treatment:"] <- NA
  }
  result$pval <- lapply(result$pval, empty_vector_if_na)
  attr(result, "conf_level") <- x$control$conf_level
  result
}

#' @describeIn tidy_coxreg Custom tidy method for a multivariate Cox regression.
#'
#' Tidy up the result of a Cox regression model fitted by [fit_coxreg_multivar()].
#'
#' @method tidy coxreg.multivar
#'
#' @examples
#' multivar_model <- fit_coxreg_multivar(
#'   variables = list(
#'     time = "time", event = "status", arm = "armcd",
#'     covariates = c("covar1", "covar2")
#'   ),
#'   data = dta_bladder
#' )
#' broom::tidy(multivar_model)
#'
#' @export
tidy.coxreg.multivar <- function(x, # nolint
                                 ...) {
  checkmate::assert_class(x, "coxreg.multivar")
  vars <- c(x$vars$arm, x$vars$covariates)

  # Convert the model summaries to data.
  result <- Map(
    vars = vars,
    f = function(vars) {
      h_coxreg_multivar_extract(
        var = vars, data = x$data,
        mod = x$mod, control = x$control
      )
    }
  )
  result <- do.call(rbind, result)

  result$ci <- Map(lcl = result$lcl, ucl = result$ucl, f = function(lcl, ucl) c(lcl, ucl))
  result$ci <- lapply(result$ci, empty_vector_if_na)
  result$hr <- lapply(result$hr, empty_vector_if_na)
  result$pval <- lapply(result$pval, empty_vector_if_na)
  result <- result[, names(result) != "n"]
  attr(result, "conf_level") <- x$control$conf_level

  result
}

#' Fitting functions for Cox proportional hazards regression
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Fitting functions for univariate and multivariate Cox regression models.
#'
#' @param variables (named `list`)\cr the names of the variables found in `data`, passed as a named list and
#'   corresponding to the `time`, `event`, `arm`, `strata`, and `covariates` terms. If `arm` is missing from
#'   `variables`, then only Cox model(s) including the `covariates` will be fitted and the corresponding effect
#'   estimates will be tabulated later.
#' @param data (`data.frame`)\cr the dataset containing the variables to fit the models.
#' @param at (`list` of `numeric`)\cr when the candidate covariate is a `numeric`, use `at` to specify
#'   the value of the covariate at which the effect should be estimated.
#' @param control (`list`)\cr a list of parameters as returned by the helper function [control_coxreg()].
#'
#' @seealso [h_cox_regression] for relevant helper functions, [cox_regression].
#'
#' @examples
#' library(survival)
#'
#' set.seed(1, kind = "Mersenne-Twister")
#'
#' # Testing dataset [survival::bladder].
#' dta_bladder <- with(
#'   data = bladder[bladder$enum < 5, ],
#'   data.frame(
#'     time = stop,
#'     status = event,
#'     armcd = as.factor(rx),
#'     covar1 = as.factor(enum),
#'     covar2 = factor(
#'       sample(as.factor(enum)),
#'       levels = 1:4, labels = c("F", "F", "M", "M")
#'     )
#'   )
#' )
#' labels <- c("armcd" = "ARM", "covar1" = "A Covariate Label", "covar2" = "Sex (F/M)")
#' formatters::var_labels(dta_bladder)[names(labels)] <- labels
#' dta_bladder$age <- sample(20:60, size = nrow(dta_bladder), replace = TRUE)
#'
#' plot(
#'   survfit(Surv(time, status) ~ armcd + covar1, data = dta_bladder),
#'   lty = 2:4,
#'   xlab = "Months",
#'   col = c("blue1", "blue2", "blue3", "blue4", "red1", "red2", "red3", "red4")
#' )
#'
#' @name fit_coxreg
NULL

#' @describeIn fit_coxreg Fit a series of univariate Cox regression models given the inputs.
#'
#' @return
#' * `fit_coxreg_univar()` returns a `coxreg.univar` class object which is a named `list`
#'   with 5 elements:
#'   * `mod`: Cox regression models fitted by [survival::coxph()].
#'   * `data`: The original data frame input.
#'   * `control`: The original control input.
#'   * `vars`: The variables used in the model.
#'   * `at`: Value of the covariate at which the effect should be estimated.
#'
#' @note When using `fit_coxreg_univar` there should be two study arms.
#'
#' @examples
#' # fit_coxreg_univar
#'
#' ## Cox regression: arm + 1 covariate.
#' mod1 <- fit_coxreg_univar(
#'   variables = list(
#'     time = "time", event = "status", arm = "armcd",
#'     covariates = "covar1"
#'   ),
#'   data = dta_bladder,
#'   control = control_coxreg(conf_level = 0.91)
#' )
#'
#' ## Cox regression: arm + 1 covariate + interaction, 2 candidate covariates.
#' mod2 <- fit_coxreg_univar(
#'   variables = list(
#'     time = "time", event = "status", arm = "armcd",
#'     covariates = c("covar1", "covar2")
#'   ),
#'   data = dta_bladder,
#'   control = control_coxreg(conf_level = 0.91, interaction = TRUE)
#' )
#'
#' ## Cox regression: arm + 1 covariate, stratified analysis.
#' mod3 <- fit_coxreg_univar(
#'   variables = list(
#'     time = "time", event = "status", arm = "armcd", strata = "covar2",
#'     covariates = c("covar1")
#'   ),
#'   data = dta_bladder,
#'   control = control_coxreg(conf_level = 0.91)
#' )
#'
#' ## Cox regression: no arm, only covariates.
#' mod4 <- fit_coxreg_univar(
#'   variables = list(
#'     time = "time", event = "status",
#'     covariates = c("covar1", "covar2")
#'   ),
#'   data = dta_bladder
#' )
#'
#' @export
fit_coxreg_univar <- function(variables,
                              data,
                              at = list(),
                              control = control_coxreg()) {
  checkmate::assert_list(variables, names = "named")
  has_arm <- "arm" %in% names(variables)
  arm_name <- if (has_arm) "arm" else NULL

  checkmate::assert_character(variables$covariates, null.ok = TRUE)

  assert_df_with_variables(data, variables)
  assert_list_of_variables(variables[c(arm_name, "event", "time")])

  if (!is.null(variables$strata)) {
    checkmate::assert_disjunct(control$pval_method, "likelihood")
  }
  if (has_arm) {
    assert_df_with_factors(data, list(val = variables$arm), min.levels = 2, max.levels = 2)
  }
  vars <- unlist(variables[c(arm_name, "covariates", "strata")], use.names = FALSE)
  for (i in vars) {
    if (is.factor(data[[i]])) {
      attr(data[[i]], "levels") <- levels(droplevels(data[[i]]))
    }
  }
  forms <- h_coxreg_univar_formulas(variables, interaction = control$interaction)
  mod <- lapply(
    forms, function(x) {
      survival::coxph(formula = stats::as.formula(x), data = data, ties = control$ties)
    }
  )
  structure(
    list(
      mod = mod,
      data = data,
      control = control,
      vars = variables,
      at = at
    ),
    class = "coxreg.univar"
  )
}

#' @describeIn fit_coxreg Fit a multivariate Cox regression model.
#'
#' @return
#' * `fit_coxreg_multivar()` returns a `coxreg.multivar` class object which is a named list
#'   with 4 elements:
#'   * `mod`: Cox regression model fitted by [survival::coxph()].
#'   * `data`: The original data frame input.
#'   * `control`: The original control input.
#'   * `vars`: The variables used in the model.
#'
#' @examples
#' # fit_coxreg_multivar
#'
#' ## Cox regression: multivariate Cox regression.
#' multivar_model <- fit_coxreg_multivar(
#'   variables = list(
#'     time = "time", event = "status", arm = "armcd",
#'     covariates = c("covar1", "covar2")
#'   ),
#'   data = dta_bladder
#' )
#'
#' # Example without treatment arm.
#' multivar_covs_model <- fit_coxreg_multivar(
#'   variables = list(
#'     time = "time", event = "status",
#'     covariates = c("covar1", "covar2")
#'   ),
#'   data = dta_bladder
#' )
#'
#' @export
fit_coxreg_multivar <- function(variables,
                                data,
                                control = control_coxreg()) {
  checkmate::assert_list(variables, names = "named")
  has_arm <- "arm" %in% names(variables)
  arm_name <- if (has_arm) "arm" else NULL

  if (!is.null(variables$covariates)) {
    checkmate::assert_character(variables$covariates)
  }

  checkmate::assert_false(control$interaction)
  assert_df_with_variables(data, variables)
  assert_list_of_variables(variables[c(arm_name, "event", "time")])

  if (!is.null(variables$strata)) {
    checkmate::assert_disjunct(control$pval_method, "likelihood")
  }

  form <- h_coxreg_multivar_formula(variables)
  mod <- survival::coxph(
    formula = stats::as.formula(form),
    data = data,
    ties = control$ties
  )
  structure(
    list(
      mod = mod,
      data = data,
      control = control,
      vars = variables
    ),
    class = "coxreg.multivar"
  )
}

#' Muffled `car::Anova`
#'
#' Applied on survival models, [car::Anova()] signal that the `strata` terms is dropped from the model formula when
#' present, this function deliberately muffles this message.
#'
#' @param mod (`coxph`)\cr Cox regression model fitted by [survival::coxph()].
#' @param test_statistic (`string`)\cr the method used for estimation of p.values; `wald` (default) or `likelihood`.
#'
#' @return The output of [car::Anova()], with convergence message muffled.
#'
#' @keywords internal
muffled_car_anova <- function(mod, test_statistic) {
  tryCatch(
    withCallingHandlers(
      expr = {
        car::Anova(
          mod,
          test.statistic = test_statistic,
          type = "III"
        )
      },
      message = function(m) invokeRestart("muffleMessage"),
      error = function(e) {
        stop(paste(
          "the model seems to have convergence problems, please try to change",
          "the configuration of covariates or strata variables, e.g.",
          "- original error:", e
        ))
      }
    )
  )
}

#' Combine factor levels
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Combine specified old factor Levels in a single new level.
#'
#' @param x (`factor`)\cr factor variable.
#' @param levels (`character`)\cr level names to be combined.
#' @param new_level (`string`)\cr name of new level.
#'
#' @return A `factor` with the new levels.
#'
#' @examples
#' x <- factor(letters[1:5], levels = letters[5:1])
#' combine_levels(x, levels = c("a", "b"))
#'
#' combine_levels(x, c("e", "b"))
#'
#' @export
combine_levels <- function(x, levels, new_level = paste(levels, collapse = "/")) {
  checkmate::assert_factor(x)
  checkmate::assert_subset(levels, levels(x))

  lvls <- levels(x)

  lvls[lvls %in% levels] <- new_level

  levels(x) <- lvls

  x
}

#' Conversion of a vector to a factor
#'
#' Converts `x` to a factor and keeps its attributes. Warns appropriately such that the user
#' can decide whether they prefer converting to factor manually (e.g. for full control of
#' factor levels).
#'
#' @param x (`vector`)\cr object to convert.
#' @param x_name (`string`)\cr name of `x`.
#' @param na_level (`string`)\cr the explicit missing level which should be used when converting a character vector.
#' @param verbose (`flag`)\cr defaults to `TRUE`. It prints out warnings and messages.
#'
#' @return A `factor` with same attributes (except class) as `x`. Does not modify `x` if already a `factor`.
#'
#' @keywords internal
as_factor_keep_attributes <- function(x,
                                      x_name = deparse(substitute(x)),
                                      na_level = "<Missing>",
                                      verbose = TRUE) {
  checkmate::assert_atomic(x)
  checkmate::assert_string(x_name)
  checkmate::assert_string(na_level)
  checkmate::assert_flag(verbose)
  if (is.factor(x)) {
    return(x)
  }
  x_class <- class(x)[1]
  if (verbose) {
    warning(paste(
      "automatically converting", x_class, "variable", x_name,
      "to factor, better manually convert to factor to avoid failures"
    ))
  }
  if (identical(length(x), 0L)) {
    warning(paste(
      x_name, "has length 0, this can lead to tabulation failures, better convert to factor"
    ))
  }
  if (is.character(x)) {
    x_no_na <- explicit_na(sas_na(x), label = na_level)
    if (any(na_level %in% x_no_na)) {
      do.call(
        structure,
        c(
          list(.Data = forcats::fct_relevel(x_no_na, na_level, after = Inf)),
          attributes(x)
        )
      )
    } else {
      do.call(structure, c(list(.Data = as.factor(x)), attributes(x)))
    }
  } else {
    do.call(structure, c(list(.Data = as.factor(x)), attributes(x)))
  }
}

#' Labels for bins in percent
#'
#' This creates labels for quantile based bins in percent. This assumes the right-closed
#' intervals as produced by [cut_quantile_bins()].
#'
#' @param probs (`numeric`)\cr the probabilities identifying the quantiles.
#'   This is a sorted vector of unique `proportion` values, i.e. between 0 and 1, where
#'   the boundaries 0 and 1 must not be included.
#' @param digits (`integer(1)`)\cr number of decimal places to round the percent numbers.
#'
#' @return A `character` vector with labels in the format `[0%,20%]`, `(20%,50%]`, etc.
#'
#' @keywords internal
bins_percent_labels <- function(probs,
                                digits = 0) {
  if (isFALSE(0 %in% probs)) probs <- c(0, probs)
  if (isFALSE(1 %in% probs)) probs <- c(probs, 1)
  checkmate::assert_numeric(probs, lower = 0, upper = 1, unique = TRUE, sorted = TRUE)
  percent <- round(probs * 100, digits = digits)
  left <- paste0(utils::head(percent, -1), "%")
  right <- paste0(utils::tail(percent, -1), "%")
  without_left_bracket <- paste0(left, ",", right, "]")
  with_left_bracket <- paste0("[", utils::head(without_left_bracket, 1))
  if (length(without_left_bracket) > 1) {
    with_left_bracket <- c(
      with_left_bracket,
      paste0("(", utils::tail(without_left_bracket, -1))
    )
  }
  with_left_bracket
}

#' Cut numeric vector into empirical quantile bins
#'
#' @description `r lifecycle::badge("stable")`
#'
#' This cuts a numeric vector into sample quantile bins.
#'
#' @inheritParams bins_percent_labels
#' @param x (`numeric`)\cr the continuous variable values which should be cut into
#'   quantile bins. This may contain `NA` values, which are then
#'   not used for the quantile calculations, but included in the return vector.
#' @param labels (`character`)\cr the unique labels for the quantile bins. When there are `n`
#'   probabilities in `probs`, then this must be `n + 1` long.
#' @param type (`integer(1)`)\cr type of quantiles to use, see [stats::quantile()] for details.
#' @param ordered (`flag`)\cr should the result be an ordered factor.
#'
#' @return A `factor` variable with appropriately-labeled bins as levels.
#'
#' @note Intervals are closed on the right side. That is, the first bin is the interval
#'   `[-Inf, q1]` where `q1` is the first quantile, the second bin is then `(q1, q2]`, etc.,
#'   and the last bin is `(qn, +Inf]` where `qn` is the last quantile.
#'
#' @examples
#' # Default is to cut into quartile bins.
#' cut_quantile_bins(cars$speed)
#'
#' # Use custom quantiles.
#' cut_quantile_bins(cars$speed, probs = c(0.1, 0.2, 0.6, 0.88))
#'
#' # Use custom labels.
#' cut_quantile_bins(cars$speed, labels = paste0("Q", 1:4))
#'
#' # NAs are preserved in result factor.
#' ozone_binned <- cut_quantile_bins(airquality$Ozone)
#' which(is.na(ozone_binned))
#' # So you might want to make these explicit.
#' explicit_na(ozone_binned)
#'
#' @export
cut_quantile_bins <- function(x,
                              probs = c(0.25, 0.5, 0.75),
                              labels = NULL,
                              type = 7,
                              ordered = TRUE) {
  checkmate::assert_flag(ordered)
  checkmate::assert_numeric(x)
  if (isFALSE(0 %in% probs)) probs <- c(0, probs)
  if (isFALSE(1 %in% probs)) probs <- c(probs, 1)
  checkmate::assert_numeric(probs, lower = 0, upper = 1, unique = TRUE, sorted = TRUE)
  if (is.null(labels)) labels <- bins_percent_labels(probs)
  checkmate::assert_character(labels, len = length(probs) - 1, any.missing = FALSE, unique = TRUE)

  if (all(is.na(x))) {
    # Early return if there are only NAs in input.
    return(factor(x, ordered = ordered, levels = labels))
  }

  quantiles <- stats::quantile(
    x,
    probs = probs,
    type = type,
    na.rm = TRUE
  )

  checkmate::assert_numeric(quantiles, unique = TRUE)

  cut(
    x,
    breaks = quantiles,
    labels = labels,
    ordered_result = ordered,
    include.lowest = TRUE,
    right = TRUE
  )
}

#' Discard specified levels of a factor
#'
#' @description `r lifecycle::badge("stable")`
#'
#' This discards the observations as well as the levels specified from a factor.
#'
#' @param x (`factor`)\cr the original factor.
#' @param discard (`character`)\cr levels to discard.
#'
#' @return A modified `factor` with observations as well as levels from `discard` dropped.
#'
#' @examples
#' fct_discard(factor(c("a", "b", "c")), "c")
#'
#' @export
fct_discard <- function(x, discard) {
  checkmate::assert_factor(x)
  checkmate::assert_character(discard, any.missing = FALSE)
  new_obs <- x[!(x %in% discard)]
  new_levels <- setdiff(levels(x), discard)
  factor(new_obs, levels = new_levels)
}

#' Insertion of explicit missing values in a factor
#'
#' @description `r lifecycle::badge("stable")`
#'
#' This inserts explicit missing values in a factor based on a condition. Additionally,
#' existing `NA` values will be explicitly converted to given `na_level`.
#'
#' @param x (`factor`)\cr the original factor.
#' @param condition (`logical`)\cr positions at which to insert missing values.
#' @param na_level (`string`)\cr which level to use for missing values.
#'
#' @return A modified `factor` with inserted and existing `NA` converted to `na_level`.
#'
#' @seealso [forcats::fct_na_value_to_level()] which is used internally.
#'
#' @examples
#' fct_explicit_na_if(factor(c("a", "b", NA)), c(TRUE, FALSE, FALSE))
#'
#' @export
fct_explicit_na_if <- function(x, condition, na_level = "<Missing>") {
  checkmate::assert_factor(x, len = length(condition))
  checkmate::assert_logical(condition)
  x[condition] <- NA
  x <- forcats::fct_na_value_to_level(x, level = na_level)
  forcats::fct_drop(x, only = na_level)
}

#' Collapse factor levels and keep only those new group levels
#'
#' @description `r lifecycle::badge("stable")`
#'
#' This collapses levels and only keeps those new group levels, in the order provided.
#' The returned factor has levels in the order given, with the possible missing level last (this will
#' only be included if there are missing values).
#'
#' @param .f (`factor` or `character`)\cr original vector.
#' @param ... (named `character`)\cr levels in each vector provided will be collapsed into
#'   the new level given by the respective name.
#' @param .na_level (`string`)\cr which level to use for other levels, which should be missing in the
#'   new factor. Note that this level must not be contained in the new levels specified in `...`.
#'
#' @return A modified `factor` with collapsed levels. Values and levels which are not included
#'   in the given `character` vector input will be set to the missing level `.na_level`.
#'
#' @note Any existing `NA`s in the input vector will not be replaced by the missing level. If needed,
#'   [explicit_na()] can be called separately on the result.
#'
#' @seealso [forcats::fct_collapse()], [forcats::fct_relevel()] which are used internally.
#'
#' @examples
#' fct_collapse_only(factor(c("a", "b", "c", "d")), TRT = "b", CTRL = c("c", "d"))
#'
#' @export
fct_collapse_only <- function(.f, ..., .na_level = "<Missing>") {
  new_lvls <- names(list(...))
  if (checkmate::test_subset(.na_level, new_lvls)) {
    stop(paste0(".na_level currently set to '", .na_level, "' must not be contained in the new levels"))
  }
  x <- forcats::fct_collapse(.f, ..., other_level = .na_level)
  do.call(forcats::fct_relevel, args = c(list(.f = x), as.list(new_lvls)))
}

#' Ungroup non-numeric statistics
#'
#' Ungroups grouped non-numeric statistics within input vectors `.formats`, `.labels`, and `.indent_mods`.
#'
#' @inheritParams argument_convention
#' @param x  (named `list` of `numeric`)\cr list of numeric statistics containing the statistics to ungroup.
#'
#' @return A `list` with modified elements `x`, `.formats`, `.labels`, and `.indent_mods`.
#'
#' @seealso [a_summary()] which uses this function internally.
#'
#' @keywords internal
ungroup_stats <- function(x,
                          .formats,
                          .labels,
                          .indent_mods) {
  checkmate::assert_list(x)
  empty_pval <- "pval" %in% names(x) && length(x[["pval"]]) == 0
  empty_pval_counts <- "pval_counts" %in% names(x) && length(x[["pval_counts"]]) == 0
  x <- unlist(x, recursive = FALSE)

  # If p-value is empty it is removed by unlist and needs to be re-added
  if (empty_pval) x[["pval"]] <- character()
  if (empty_pval_counts) x[["pval_counts"]] <- character()
  .stats <- names(x)

  # Ungroup stats
  .formats <- lapply(.stats, function(x) {
    .formats[[if (!grepl("\\.", x)) x else regmatches(x, regexpr("\\.", x), invert = TRUE)[[1]][1]]]
  })
  .indent_mods <- sapply(.stats, function(x) {
    .indent_mods[[if (!grepl("\\.", x)) x else regmatches(x, regexpr("\\.", x), invert = TRUE)[[1]][1]]]
  })
  .labels <- sapply(.stats, function(x) {
    if (!grepl("\\.", x)) .labels[[x]] else regmatches(x, regexpr("\\.", x), invert = TRUE)[[1]][2]
  })

  list(
    x = x,
    .formats = .formats,
    .labels = .labels,
    .indent_mods = .indent_mods
  )
}

#' Tabulate binary response by subgroup
#'
#' @description `r lifecycle::badge("stable")`
#'
#' The [tabulate_rsp_subgroups()] function creates a layout element to tabulate binary response by subgroup, returning
#' statistics including response rate and odds ratio for each population subgroup. The table is created from `df`, a
#' list of data frames returned by [extract_rsp_subgroups()], with the statistics to include specified via the `vars`
#' parameter.
#'
#' A forest plot can be created from the resulting table using the [g_forest()] function.
#'
#' @inheritParams extract_rsp_subgroups
#' @inheritParams argument_convention
#'
#' @details These functions create a layout starting from a data frame which contains
#'   the required statistics. Tables typically used as part of forest plot.
#'
#' @seealso [extract_rsp_subgroups()]
#'
#' @examples
#' library(dplyr)
#' library(forcats)
#'
#' adrs <- tern_ex_adrs
#' adrs_labels <- formatters::var_labels(adrs)
#'
#' adrs_f <- adrs %>%
#'   filter(PARAMCD == "BESRSPI") %>%
#'   filter(ARM %in% c("A: Drug X", "B: Placebo")) %>%
#'   droplevels() %>%
#'   mutate(
#'     # Reorder levels of factor to make the placebo group the reference arm.
#'     ARM = fct_relevel(ARM, "B: Placebo"),
#'     rsp = AVALC == "CR"
#'   )
#' formatters::var_labels(adrs_f) <- c(adrs_labels, "Response")
#'
#' # Unstratified analysis.
#' df <- extract_rsp_subgroups(
#'   variables = list(rsp = "rsp", arm = "ARM", subgroups = c("SEX", "BMRKR2")),
#'   data = adrs_f
#' )
#' df
#'
#' # Stratified analysis.
#' df_strat <- extract_rsp_subgroups(
#'   variables = list(rsp = "rsp", arm = "ARM", subgroups = c("SEX", "BMRKR2"), strata = "STRATA1"),
#'   data = adrs_f
#' )
#' df_strat
#'
#' # Grouping of the BMRKR2 levels.
#' df_grouped <- extract_rsp_subgroups(
#'   variables = list(rsp = "rsp", arm = "ARM", subgroups = c("SEX", "BMRKR2")),
#'   data = adrs_f,
#'   groups_lists = list(
#'     BMRKR2 = list(
#'       "low" = "LOW",
#'       "low/medium" = c("LOW", "MEDIUM"),
#'       "low/medium/high" = c("LOW", "MEDIUM", "HIGH")
#'     )
#'   )
#' )
#' df_grouped
#'
#' @name response_subgroups
#' @order 1
NULL

#' Prepare response data for population subgroups in data frames
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Prepares response rates and odds ratios for population subgroups in data frames. Simple wrapper
#' for [h_odds_ratio_subgroups_df()] and [h_proportion_subgroups_df()]. Result is a list of two
#' `data.frames`: `prop` and `or`. `variables` corresponds to the names of variables found in `data`,
#' passed as a named `list` and requires elements `rsp`, `arm` and optionally `subgroups` and `strata`.
#' `groups_lists` optionally specifies groupings for `subgroups` variables.
#'
#' @inheritParams argument_convention
#' @inheritParams response_subgroups
#' @param label_all (`string`)\cr label for the total population analysis.
#'
#' @return A named list of two elements:
#'   * `prop`: A `data.frame` containing columns `arm`, `n`, `n_rsp`, `prop`, `subgroup`, `var`,
#'     `var_label`, and `row_type`.
#'   * `or`: A `data.frame` containing columns `arm`, `n_tot`, `or`, `lcl`, `ucl`, `conf_level`,
#'     `subgroup`, `var`, `var_label`, and `row_type`.
#'
#' @seealso [response_subgroups]
#'
#' @export
extract_rsp_subgroups <- function(variables,
                                  data,
                                  groups_lists = list(),
                                  conf_level = 0.95,
                                  method = NULL,
                                  label_all = "All Patients") {
  if ("strat" %in% names(variables)) {
    warning(
      "Warning: the `strat` element name of the `variables` list argument to `extract_rsp_subgroups() ",
      "was deprecated in tern 0.9.4.\n  ",
      "Please use the name `strata` instead of `strat` in the `variables` argument."
    )
    variables[["strata"]] <- variables[["strat"]]
  }

  df_prop <- h_proportion_subgroups_df(
    variables,
    data,
    groups_lists = groups_lists,
    label_all = label_all
  )
  df_or <- h_odds_ratio_subgroups_df(
    variables,
    data,
    groups_lists = groups_lists,
    conf_level = conf_level,
    method = method,
    label_all = label_all
  )

  list(prop = df_prop, or = df_or)
}

#' @describeIn response_subgroups Formatted analysis function which is used as `afun` in `tabulate_rsp_subgroups()`.
#'
#' @return
#' * `a_response_subgroups()` returns the corresponding list with formatted [rtables::CellValue()].
#'
#' @keywords internal
a_response_subgroups <- function(.formats = list(
                                   n = "xx", # nolint start
                                   n_rsp = "xx",
                                   prop = "xx.x%",
                                   n_tot = "xx",
                                   or = list(format_extreme_values(2L)),
                                   ci = list(format_extreme_values_ci(2L)),
                                   pval = "x.xxxx | (<0.0001)",
                                   riskdiff = "xx.x (xx.x - xx.x)" # nolint end
                                 ),
                                 na_str = default_na_str()) {
  checkmate::assert_list(.formats)
  checkmate::assert_subset(
    names(.formats),
    c("n", "n_rsp", "prop", "n_tot", "or", "ci", "pval", "riskdiff")
  )

  afun_lst <- Map(
    function(stat, fmt, na_str) {
      function(df, labelstr = "", ...) {
        in_rows(
          .list = as.list(df[[stat]]),
          .labels = as.character(df$subgroup),
          .formats = fmt,
          .format_na_strs = na_str
        )
      }
    },
    stat = names(.formats),
    fmt = .formats,
    na_str = na_str
  )

  afun_lst
}

#' @describeIn response_subgroups Table-creating function which creates a table
#'   summarizing binary response by subgroup. This function is a wrapper for [rtables::analyze_colvars()]
#'   and [rtables::summarize_row_groups()].
#'
#' @param df (`list`)\cr a list of data frames containing all analysis variables. List should be
#'   created using [extract_rsp_subgroups()].
#' @param vars (`character`)\cr the names of statistics to be reported among:
#'   * `n`: Total number of observations per group.
#'   * `n_rsp`: Number of responders per group.
#'   * `prop`: Proportion of responders.
#'   * `n_tot`: Total number of observations.
#'   * `or`: Odds ratio.
#'   * `ci` : Confidence interval of odds ratio.
#'   * `pval`: p-value of the effect.
#'   Note, the statistics `n_tot`, `or`, and `ci` are required.
#' @param riskdiff (`list`)\cr if a risk (proportion) difference column should be added, a list of settings to apply
#'   within the column. See [control_riskdiff()] for details. If `NULL`, no risk difference column will be added. If
#'   `riskdiff$arm_x` and `riskdiff$arm_y` are `NULL`, the first level of `df$prop$arm` will be used as `arm_x` and
#'   the second level as `arm_y`.
#'
#' @return An `rtables` table summarizing binary response by subgroup.
#'
#' @examples
#' # Table with default columns
#' basic_table() %>%
#'   tabulate_rsp_subgroups(df)
#'
#' # Table with selected columns
#' basic_table() %>%
#'   tabulate_rsp_subgroups(
#'     df = df,
#'     vars = c("n_tot", "n", "n_rsp", "prop", "or", "ci")
#'   )
#'
#' # Table with risk difference column added
#' basic_table() %>%
#'   tabulate_rsp_subgroups(
#'     df,
#'     riskdiff = control_riskdiff(
#'       arm_x = levels(df$prop$arm)[1],
#'       arm_y = levels(df$prop$arm)[2]
#'     )
#'   )
#'
#' @export
#' @order 2
tabulate_rsp_subgroups <- function(lyt,
                                   df,
                                   vars = c("n_tot", "n", "prop", "or", "ci"),
                                   groups_lists = list(),
                                   label_all = "All Patients",
                                   riskdiff = NULL,
                                   na_str = default_na_str(),
                                   .formats = c(
                                     n = "xx", n_rsp = "xx", prop = "xx.x%", n_tot = "xx",
                                     or = list(format_extreme_values(2L)), ci = list(format_extreme_values_ci(2L)),
                                     pval = "x.xxxx | (<0.0001)"
                                   )) {
  checkmate::assert_list(riskdiff, null.ok = TRUE)
  checkmate::assert_true(all(c("n_tot", "or", "ci") %in% vars))
  if ("pval" %in% vars && !"pval" %in% names(df$or)) {
    warning(
      'The "pval" statistic has been selected but is not present in "df" so it will not be included in the output ',
      'table. To include the "pval" statistic, please specify a p-value test when generating "df" via ',
      'the "method" argument to `extract_rsp_subgroups()`. If method = "cmh", strata must also be specified via the ',
      '"variables" argument to `extract_rsp_subgroups()`.'
    )
  }

  # Create "ci" column from "lcl" and "ucl"
  df$or$ci <- combine_vectors(df$or$lcl, df$or$ucl)

  # Fill in missing formats with defaults
  default_fmts <- eval(formals(tabulate_rsp_subgroups)$.formats)
  .formats <- c(.formats, default_fmts[vars[!vars %in% names(.formats)]])

  # Extract additional parameters from df
  conf_level <- df$or$conf_level[1]
  method <- if ("pval_label" %in% names(df$or)) df$or$pval_label[1] else NULL
  colvars <- d_rsp_subgroups_colvars(vars, conf_level = conf_level, method = method)
  prop_vars <- intersect(colvars$vars, c("n", "prop", "n_rsp"))
  or_vars <- intersect(names(colvars$labels), c("n_tot", "or", "ci", "pval"))
  colvars_prop <- list(vars = prop_vars, labels = colvars$labels[prop_vars])
  colvars_or <- list(vars = or_vars, labels = colvars$labels[or_vars])

  extra_args <- list(groups_lists = groups_lists, conf_level = conf_level, method = method, label_all = label_all)

  # Get analysis function for each statistic
  afun_lst <- a_response_subgroups(.formats = c(.formats, riskdiff = riskdiff$format), na_str = na_str)

  # Add risk difference column
  if (!is.null(riskdiff)) {
    if (is.null(riskdiff$arm_x)) riskdiff$arm_x <- levels(df$prop$arm)[1]
    if (is.null(riskdiff$arm_y)) riskdiff$arm_y <- levels(df$prop$arm)[2]
    colvars_or$vars <- c(colvars_or$vars, "riskdiff")
    colvars_or$labels <- c(colvars_or$labels, riskdiff = riskdiff$col_label)
    arm_cols <- paste(rep(c("n_rsp", "n_rsp", "n", "n")), c(riskdiff$arm_x, riskdiff$arm_y), sep = "_")

    df_prop_diff <- df$prop %>%
      dplyr::select(-"prop") %>%
      tidyr::pivot_wider(
        id_cols = c("subgroup", "var", "var_label", "row_type"),
        names_from = "arm",
        values_from = c("n", "n_rsp")
      ) %>%
      dplyr::rowwise() %>%
      dplyr::mutate(
        riskdiff = stat_propdiff_ci(
          x = as.list(.data[[arm_cols[1]]]),
          y = as.list(.data[[arm_cols[2]]]),
          N_x = .data[[arm_cols[3]]],
          N_y = .data[[arm_cols[4]]]
        )
      ) %>%
      dplyr::select(-dplyr::all_of(arm_cols))

    df$or <- df$or %>%
      dplyr::left_join(
        df_prop_diff,
        by = c("subgroup", "var", "var_label", "row_type")
      )
  }

  # Add columns from table_prop (optional)
  if (length(colvars_prop$vars) > 0) {
    lyt_prop <- split_cols_by(lyt = lyt, var = "arm")
    lyt_prop <- split_cols_by_multivar(
      lyt = lyt_prop,
      vars = colvars_prop$vars,
      varlabels = colvars_prop$labels
    )

    # Add "All Patients" row
    lyt_prop <- split_rows_by(
      lyt = lyt_prop,
      var = "row_type",
      split_fun = keep_split_levels("content"),
      nested = FALSE,
      child_labels = "hidden"
    )
    lyt_prop <- analyze_colvars(
      lyt = lyt_prop,
      afun = afun_lst[names(colvars_prop$labels)],
      na_str = na_str,
      extra_args = extra_args
    )

    # Add analysis rows
    if ("analysis" %in% df$prop$row_type) {
      lyt_prop <- split_rows_by(
        lyt = lyt_prop,
        var = "row_type",
        split_fun = keep_split_levels("analysis"),
        nested = FALSE,
        child_labels = "hidden"
      )
      lyt_prop <- split_rows_by(lyt = lyt_prop, var = "var_label", nested = TRUE)
      lyt_prop <- analyze_colvars(
        lyt = lyt_prop,
        afun = afun_lst[names(colvars_prop$labels)],
        na_str = na_str,
        inclNAs = TRUE,
        extra_args = extra_args
      )
    }

    table_prop <- build_table(lyt_prop, df = df$prop)
  } else {
    table_prop <- NULL
  }

  # Add columns from table_or ("n_tot", "or", and "ci" required)
  lyt_or <- split_cols_by(lyt = lyt, var = "arm")
  lyt_or <- split_cols_by_multivar(
    lyt = lyt_or,
    vars = colvars_or$vars,
    varlabels = colvars_or$labels
  )

  # Add "All Patients" row
  lyt_or <- split_rows_by(
    lyt = lyt_or,
    var = "row_type",
    split_fun = keep_split_levels("content"),
    nested = FALSE,
    child_labels = "hidden"
  )
  lyt_or <- analyze_colvars(
    lyt = lyt_or,
    afun = afun_lst[names(colvars_or$labels)],
    na_str = na_str,
    extra_args = extra_args
  ) %>%
    append_topleft("Baseline Risk Factors")

  # Add analysis rows
  if ("analysis" %in% df$or$row_type) {
    lyt_or <- split_rows_by(
      lyt = lyt_or,
      var = "row_type",
      split_fun = keep_split_levels("analysis"),
      nested = FALSE,
      child_labels = "hidden"
    )
    lyt_or <- split_rows_by(lyt = lyt_or, var = "var_label", nested = TRUE)
    lyt_or <- analyze_colvars(
      lyt = lyt_or,
      afun = afun_lst[names(colvars_or$labels)],
      na_str = na_str,
      inclNAs = TRUE,
      extra_args = extra_args
    )
  }

  table_or <- build_table(lyt_or, df = df$or)

  # Join tables, add forest plot attributes
  n_tot_id <- match("n_tot", colvars_or$vars)
  if (is.null(table_prop)) {
    result <- table_or
    or_id <- match("or", colvars_or$vars)
    ci_id <- match("ci", colvars_or$vars)
  } else {
    result <- cbind_rtables(table_or[, n_tot_id], table_prop, table_or[, -n_tot_id])
    or_id <- 1L + ncol(table_prop) + match("or", colvars_or$vars[-n_tot_id])
    ci_id <- 1L + ncol(table_prop) + match("ci", colvars_or$vars[-n_tot_id])
    n_tot_id <- 1L
  }
  structure(
    result,
    forest_header = paste0(levels(df$prop$arm), "\nBetter"),
    col_x = or_id,
    col_ci = ci_id,
    col_symbol_size = n_tot_id
  )
}

#' Labels for column variables in binary response by subgroup table
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Internal function to check variables included in [tabulate_rsp_subgroups()] and create column labels.
#'
#' @inheritParams argument_convention
#' @inheritParams tabulate_rsp_subgroups
#'
#' @return A `list` of variables to tabulate and their labels.
#'
#' @export
d_rsp_subgroups_colvars <- function(vars,
                                    conf_level = NULL,
                                    method = NULL) {
  checkmate::assert_character(vars)
  checkmate::assert_subset(c("n_tot", "or", "ci"), vars)
  checkmate::assert_subset(
    vars,
    c("n", "n_rsp", "prop", "n_tot", "or", "ci", "pval")
  )

  varlabels <- c(
    n = "n",
    n_rsp = "Responders",
    prop = "Response (%)",
    n_tot = "Total n",
    or = "Odds Ratio"
  )
  colvars <- vars

  if ("ci" %in% colvars) {
    checkmate::assert_false(is.null(conf_level))

    varlabels <- c(
      varlabels,
      ci = paste0(100 * conf_level, "% CI")
    )

    # The `lcl`` variable is just a placeholder available in the analysis data,
    # it is not acutally used in the tabulation.
    # Variables used in the tabulation are lcl and ucl, see `a_response_subgroups` for details.
    colvars[colvars == "ci"] <- "lcl"
  }

  if ("pval" %in% colvars) {
    varlabels <- c(
      varlabels,
      pval = method
    )
  }

  list(
    vars = colvars,
    labels = varlabels[vars]
  )
}

#' Get default statistical methods and their associated formats, labels, and indent modifiers
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Utility functions to get valid statistic methods for different method groups
#' (`.stats`) and their associated formats (`.formats`), labels (`.labels`), and indent modifiers
#' (`.indent_mods`). This utility is used across `tern`, but some of its working principles can be
#' seen in [analyze_vars()]. See notes to understand why this is experimental.
#'
#' @param stats (`character`)\cr statistical methods to get defaults for.
#'
#' @details
#' Current choices for `type` are `counts` and `numeric` for [analyze_vars()] and affect `get_stats()`.
#'
#' @note
#' These defaults are experimental because we use the names of functions to retrieve the default
#' statistics. This should be generalized in groups of methods according to more reasonable groupings.
#'
#' @name default_stats_formats_labels
NULL

#' @describeIn default_stats_formats_labels Get statistics available for a given method
#'   group (analyze function). To check available defaults see `tern::tern_default_stats` list.
#'
#' @param method_groups (`character`)\cr indicates the statistical method group (`tern` analyze function)
#'   to retrieve default statistics for. A character vector can be used to specify more than one statistical
#'   method group.
#' @param stats_in (`character`)\cr statistics to retrieve for the selected method group.
#' @param add_pval (`flag`)\cr should `"pval"` (or `"pval_counts"` if `method_groups` contains
#'   `"analyze_vars_counts"`) be added to the statistical methods?
#'
#' @return
#' * `get_stats()` returns a `character` vector of statistical methods.
#'
#' @examples
#' # analyze_vars is numeric
#' num_stats <- get_stats("analyze_vars_numeric") # also the default
#'
#' # Other type
#' cnt_stats <- get_stats("analyze_vars_counts")
#'
#' # Weirdly taking the pval from count_occurrences
#' only_pval <- get_stats("count_occurrences", add_pval = TRUE, stats_in = "pval")
#'
#' # All count_occurrences
#' all_cnt_occ <- get_stats("count_occurrences")
#'
#' # Multiple
#' get_stats(c("count_occurrences", "analyze_vars_counts"))
#'
#' @export
get_stats <- function(method_groups = "analyze_vars_numeric", stats_in = NULL, add_pval = FALSE) {
  checkmate::assert_character(method_groups)
  checkmate::assert_character(stats_in, null.ok = TRUE)
  checkmate::assert_flag(add_pval)

  # Default is still numeric
  if (any(method_groups == "analyze_vars")) {
    method_groups[method_groups == "analyze_vars"] <- "analyze_vars_numeric"
  }

  type_tmp <- ifelse(any(grepl("counts", method_groups)), "counts", "numeric") # for pval checks

  # Defaults for loop
  out <- NULL

  # Loop for multiple method groups
  for (mgi in method_groups) {
    out_tmp <- if (mgi %in% names(tern_default_stats)) {
      tern_default_stats[[mgi]]
    } else {
      stop("The selected method group (", mgi, ") has no default statistical method.")
    }
    out <- unique(c(out, out_tmp))
  }

  # If you added pval to the stats_in you certainly want it
  if (!is.null(stats_in) && any(grepl("^pval", stats_in))) {
    stats_in_pval_value <- stats_in[grepl("^pval", stats_in)]

    # Must be only one value between choices
    checkmate::assert_choice(stats_in_pval_value, c("pval", "pval_counts"))

    # Mismatch with counts and numeric
    if (any(grepl("counts", method_groups)) && stats_in_pval_value != "pval_counts" ||
      any(grepl("numeric", method_groups)) && stats_in_pval_value != "pval") { # nolint
      stop(
        "Inserted p-value (", stats_in_pval_value, ") is not valid for type ",
        type_tmp, ". Use ", paste(ifelse(stats_in_pval_value == "pval", "pval_counts", "pval")),
        " instead."
      )
    }

    # Lets add it even if present (thanks to unique)
    add_pval <- TRUE
  }

  # Mainly used in "analyze_vars" but it could be necessary elsewhere
  if (isTRUE(add_pval)) {
    if (any(grepl("counts", method_groups))) {
      out <- unique(c(out, "pval_counts"))
    } else {
      out <- unique(c(out, "pval"))
    }
  }

  # Filtering for stats_in (character vector)
  if (!is.null(stats_in)) {
    out <- intersect(stats_in, out) # It orders them too
  }

  # If intersect did not find matches (and no pval?) -> error
  if (length(out) == 0) {
    stop(
      "The selected method group(s) (", paste0(method_groups, collapse = ", "), ")",
      " do not have the required default statistical methods:\n",
      paste0(stats_in, collapse = " ")
    )
  }

  out
}

#' @describeIn default_stats_formats_labels Get formats corresponding to a list of statistics.
#'   To check available defaults see `tern::tern_default_formats` list.
#'
#' @param formats_in (named `vector`)\cr inserted formats to replace defaults. It can be a
#'   character vector from [formatters::list_valid_format_labels()] or a custom format function.
#'
#' @return
#' * `get_formats_from_stats()` returns a named vector of formats (if present in either
#'   `tern_default_formats` or `formats_in`, otherwise `NULL`). Values can be taken from
#'   [formatters::list_valid_format_labels()] or a custom function (e.g. [formatting_functions]).
#'
#' @note Formats in `tern` and `rtables` can be functions that take in the table cell value and
#'   return a string. This is well documented in `vignette("custom_appearance", package = "rtables")`.
#'
#' @examples
#' # Defaults formats
#' get_formats_from_stats(num_stats)
#' get_formats_from_stats(cnt_stats)
#' get_formats_from_stats(only_pval)
#' get_formats_from_stats(all_cnt_occ)
#'
#' # Addition of customs
#' get_formats_from_stats(all_cnt_occ, formats_in = c("fraction" = c("xx")))
#' get_formats_from_stats(all_cnt_occ, formats_in = list("fraction" = c("xx.xx", "xx")))
#'
#' @seealso [formatting_functions]
#'
#' @export
get_formats_from_stats <- function(stats, formats_in = NULL) {
  checkmate::assert_character(stats, min.len = 1)
  # It may be a list if there is a function in the formats
  if (checkmate::test_list(formats_in, null.ok = TRUE)) {
    checkmate::assert_list(formats_in, null.ok = TRUE)
    # Or it may be a vector of characters
  } else {
    checkmate::assert_character(formats_in, null.ok = TRUE)
  }

  # Extract global defaults
  which_fmt <- match(stats, names(tern_default_formats))

  # Select only needed formats from stats
  ret <- vector("list", length = length(stats)) # Returning a list is simpler
  ret[!is.na(which_fmt)] <- tern_default_formats[which_fmt[!is.na(which_fmt)]]

  out <- setNames(ret, stats)

  # Modify some with custom formats
  if (!is.null(formats_in)) {
    # Stats is the main
    common_names <- intersect(names(out), names(formats_in))
    out[common_names] <- formats_in[common_names]
  }

  out
}

#' @describeIn default_stats_formats_labels Get labels corresponding to a list of statistics.
#'   To check for available defaults see `tern::tern_default_labels` list. If not available there,
#'   the statistics name will be used as label.
#'
#' @param labels_in (named `character`)\cr inserted labels to replace defaults.
#' @param row_nms (`character`)\cr row names. Levels of a `factor` or `character` variable, each
#'   of which the statistics in `.stats` will be calculated for. If this parameter is set, these
#'   variable levels will be used as the defaults, and the names of the given custom values should
#'   correspond to levels (or have format `statistic.level`) instead of statistics. Can also be
#'   variable names if rows correspond to different variables instead of levels. Defaults to `NULL`.
#'
#' @return
#' * `get_labels_from_stats()` returns a named `character` vector of labels (if present in either
#'   `tern_default_labels` or `labels_in`, otherwise `NULL`).
#'
#' @examples
#' # Defaults labels
#' get_labels_from_stats(num_stats)
#' get_labels_from_stats(cnt_stats)
#' get_labels_from_stats(only_pval)
#' get_labels_from_stats(all_cnt_occ)
#'
#' # Addition of customs
#' get_labels_from_stats(all_cnt_occ, labels_in = c("fraction" = "Fraction"))
#' get_labels_from_stats(all_cnt_occ, labels_in = list("fraction" = c("Some more fractions")))
#'
#' @export
get_labels_from_stats <- function(stats, labels_in = NULL, row_nms = NULL) {
  checkmate::assert_character(stats, min.len = 1)
  checkmate::assert_character(row_nms, null.ok = TRUE)
  # It may be a list
  if (checkmate::test_list(labels_in, null.ok = TRUE)) {
    checkmate::assert_list(labels_in, null.ok = TRUE)
    # Or it may be a vector of characters
  } else {
    checkmate::assert_character(labels_in, null.ok = TRUE)
  }

  if (!is.null(row_nms)) {
    ret <- rep(row_nms, length(stats))
    out <- setNames(ret, paste(rep(stats, each = length(row_nms)), ret, sep = "."))

    if (!is.null(labels_in)) {
      lvl_lbls <- intersect(names(labels_in), row_nms)
      for (i in lvl_lbls) out[paste(stats, i, sep = ".")] <- labels_in[[i]]
    }
  } else {
    which_lbl <- match(stats, names(tern_default_labels))

    ret <- stats # The default
    ret[!is.na(which_lbl)] <- tern_default_labels[which_lbl[!is.na(which_lbl)]]

    out <- setNames(ret, stats)
  }

  # Modify some with custom labels
  if (!is.null(labels_in)) {
    # Stats is the main
    common_names <- intersect(names(out), names(labels_in))
    out[common_names] <- labels_in[common_names]
  }

  out
}

#' @describeIn default_stats_formats_labels Format indent modifiers for a given vector/list of statistics.
#'   It defaults to 0L for all values.
#'
#' @param indents_in (named `vector`)\cr inserted indent modifiers to replace defaults (default is `0L`).
#'
#' @return
#' * `get_indents_from_stats()` returns a single indent modifier value to apply to all rows
#'   or a named numeric vector of indent modifiers (if present, otherwise `NULL`).
#'
#' @examples
#' get_indents_from_stats(all_cnt_occ, indents_in = 3L)
#' get_indents_from_stats(all_cnt_occ, indents_in = list(count = 2L, count_fraction = 5L))
#' get_indents_from_stats(
#'   all_cnt_occ,
#'   indents_in = list(a = 2L, count.a = 1L, count.b = 5L), row_nms = c("a", "b")
#' )
#'
#' @export
get_indents_from_stats <- function(stats, indents_in = NULL, row_nms = NULL) {
  checkmate::assert_character(stats, min.len = 1)
  checkmate::assert_character(row_nms, null.ok = TRUE)
  # It may be a list
  if (checkmate::test_list(indents_in, null.ok = TRUE)) {
    checkmate::assert_list(indents_in, null.ok = TRUE)
    # Or it may be a vector of integers
  } else {
    checkmate::assert_integerish(indents_in, null.ok = TRUE)
  }

  if (is.null(names(indents_in)) && length(indents_in) == 1) {
    out <- rep(indents_in, length(stats) * if (!is.null(row_nms)) length(row_nms) else 1)
    return(out)
  }

  if (!is.null(row_nms)) {
    ret <- rep(0L, length(stats) * length(row_nms))
    out <- setNames(ret, paste(rep(stats, each = length(row_nms)), rep(row_nms, length(stats)), sep = "."))

    if (!is.null(indents_in)) {
      lvl_lbls <- intersect(names(indents_in), row_nms)
      for (i in lvl_lbls) out[paste(stats, i, sep = ".")] <- indents_in[[i]]
    }
  } else {
    ret <- rep(0L, length(stats))
    out <- setNames(ret, stats)
  }

  # Modify some with custom labels
  if (!is.null(indents_in)) {
    # Stats is the main
    common_names <- intersect(names(out), names(indents_in))
    out[common_names] <- indents_in[common_names]
  }

  out
}

#' Update labels according to control specifications
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Given a list of statistic labels and and a list of control parameters, updates labels with a relevant
#' control specification. For example, if control has element `conf_level` set to `0.9`, the default
#' label for statistic `mean_ci` will be updated to `"Mean 90% CI"`. Any labels that are supplied
#' via `labels_custom` will not be updated regardless of `control`.
#'
#' @param labels_default (named `character`)\cr a named vector of statistic labels to modify
#'   according to the control specifications. Labels that are explicitly defined in `labels_custom` will
#'   not be affected.
#' @param labels_custom (named `character`)\cr named vector of labels that are customized by
#'   the user and should not be affected by `control`.
#' @param control (named `list`)\cr list of control parameters to apply to adjust default labels.
#'
#' @return A named character vector of labels with control specifications applied to relevant labels.
#'
#' @examples
#' control <- list(conf_level = 0.80, quantiles = c(0.1, 0.83), test_mean = 0.57)
#' get_labels_from_stats(c("mean_ci", "quantiles", "mean_pval")) %>%
#'   labels_use_control(control = control)
#'
#' @export
labels_use_control <- function(labels_default, control, labels_custom = NULL) {
  if ("conf_level" %in% names(control)) {
    labels_default <- sapply(
      names(labels_default),
      function(x) {
        if (!x %in% names(labels_custom)) {
          gsub(labels_default[[x]], pattern = "[0-9]+% CI", replacement = f_conf_level(control[["conf_level"]]))
        } else {
          labels_default[[x]]
        }
      }
    )
  }
  if ("quantiles" %in% names(control) && "quantiles" %in% names(labels_default) &&
    !"quantiles" %in% names(labels_custom)) { # nolint
    labels_default["quantiles"] <- gsub(
      "[0-9]+% and [0-9]+", paste0(control[["quantiles"]][1] * 100, "% and ", control[["quantiles"]][2] * 100, ""),
      labels_default["quantiles"]
    )
  }
  if ("test_mean" %in% names(control) && "mean_pval" %in% names(labels_default) &&
    !"mean_pval" %in% names(labels_custom)) { # nolint
    labels_default["mean_pval"] <- gsub(
      "p-value \\(H0: mean = [0-9\\.]+\\)", f_pval(control[["test_mean"]]), labels_default["mean_pval"]
    )
  }

  labels_default
}

#' @describeIn default_stats_formats_labels Named list of available statistics by method group for `tern`.
#'
#' @format
#' * `tern_default_stats` is a named list of available statistics, with each element
#'   named for their corresponding statistical method group.
#'
#' @export
tern_default_stats <- list(
  abnormal = c("fraction"),
  abnormal_by_baseline = c("fraction"),
  abnormal_by_marked = c("count_fraction", "count_fraction_fixed_dp"),
  abnormal_by_worst_grade = c("count_fraction", "count_fraction_fixed_dp"),
  abnormal_by_worst_grade_worsen = c("fraction"),
  analyze_patients_exposure_in_cols = c("n_patients", "sum_exposure"),
  analyze_vars_counts = c("n", "count", "count_fraction", "count_fraction_fixed_dp", "fraction", "n_blq"),
  analyze_vars_numeric = c(
    "n", "sum", "mean", "sd", "se", "mean_sd", "mean_se", "mean_ci", "mean_sei", "mean_sdi", "mean_pval",
    "median", "mad", "median_ci", "quantiles", "iqr", "range", "min", "max", "median_range", "cv",
    "geom_mean", "geom_mean_ci", "geom_cv"
  ),
  count_cumulative = c("count_fraction", "count_fraction_fixed_dp"),
  count_missed_doses = c("n", "count_fraction", "count_fraction_fixed_dp"),
  count_occurrences = c("count", "count_fraction", "count_fraction_fixed_dp", "fraction"),
  count_occurrences_by_grade = c("count_fraction", "count_fraction_fixed_dp"),
  count_patients_with_event = c("n", "count", "count_fraction", "count_fraction_fixed_dp", "n_blq"),
  count_patients_with_flags = c("n", "count", "count_fraction", "count_fraction_fixed_dp", "n_blq"),
  count_values = c("n", "count", "count_fraction", "count_fraction_fixed_dp", "n_blq"),
  coxph_pairwise = c("pvalue", "hr", "hr_ci", "n_tot", "n_tot_events"),
  estimate_incidence_rate = c("person_years", "n_events", "rate", "rate_ci", "n_unique", "n_rate"),
  estimate_multinomial_response = c("n_prop", "prop_ci"),
  estimate_odds_ratio = c("or_ci", "n_tot"),
  estimate_proportion = c("n_prop", "prop_ci"),
  estimate_proportion_diff = c("diff", "diff_ci"),
  summarize_ancova = c("n", "lsmean", "lsmean_diff", "lsmean_diff_ci", "pval"),
  summarize_coxreg = c("n", "hr", "ci", "pval", "pval_inter"),
  summarize_glm_count = c("n", "rate", "rate_ci", "rate_ratio", "rate_ratio_ci", "pval"),
  summarize_num_patients = c("unique", "nonunique", "unique_count"),
  summarize_patients_events_in_cols = c("unique", "all"),
  surv_time = c("median", "median_ci", "quantiles", "range_censor", "range_event", "range"),
  surv_timepoint = c("pt_at_risk", "event_free_rate", "rate_se", "rate_ci", "rate_diff", "rate_diff_ci", "ztest_pval"),
  tabulate_rsp_biomarkers = c("n_tot", "n_rsp", "prop", "or", "ci", "pval"),
  tabulate_rsp_subgroups = c("n", "n_rsp", "prop", "n_tot", "or", "ci", "pval"),
  tabulate_survival_biomarkers = c("n_tot", "n_tot_events", "median", "hr", "ci", "pval"),
  tabulate_survival_subgroups = c("n_tot_events", "n_events", "n_tot", "n", "median", "hr", "ci", "pval"),
  test_proportion_diff = c("pval")
)

#' @describeIn default_stats_formats_labels Named vector of default formats for `tern`.
#'
#' @format
#' * `tern_default_formats` is a named vector of available default formats, with each element
#'   named for their corresponding statistic.
#'
#' @export
tern_default_formats <- c(
  fraction = format_fraction_fixed_dp,
  unique = format_count_fraction_fixed_dp,
  nonunique = "xx",
  unique_count = "xx",
  n = "xx.",
  count = "xx.",
  count_fraction = format_count_fraction,
  count_fraction_fixed_dp = format_count_fraction_fixed_dp,
  n_blq = "xx.",
  sum = "xx.x",
  mean = "xx.x",
  sd = "xx.x",
  se = "xx.x",
  mean_sd = "xx.x (xx.x)",
  mean_se = "xx.x (xx.x)",
  mean_ci = "(xx.xx, xx.xx)",
  mean_sei = "(xx.xx, xx.xx)",
  mean_sdi = "(xx.xx, xx.xx)",
  mean_pval = "x.xxxx | (<0.0001)",
  median = "xx.x",
  mad = "xx.x",
  median_ci = "(xx.xx, xx.xx)",
  quantiles = "xx.x - xx.x",
  iqr = "xx.x",
  range = "xx.x - xx.x",
  min = "xx.x",
  max = "xx.x",
  median_range = "xx.x (xx.x - xx.x)",
  cv = "xx.x",
  geom_mean = "xx.x",
  geom_mean_ci = "(xx.xx, xx.xx)",
  geom_cv = "xx.x",
  pval = "x.xxxx | (<0.0001)",
  pval_counts = "x.xxxx | (<0.0001)",
  range_censor = "xx.x to xx.x",
  range_event = "xx.x to xx.x",
  rate = "xx.xxxx",
  rate_ci = "(xx.xxxx, xx.xxxx)",
  rate_ratio = "xx.xxxx",
  rate_ratio_ci = "(xx.xxxx, xx.xxxx)"
)

#' @describeIn default_stats_formats_labels Named `character` vector of default labels for `tern`.
#'
#' @format
#' * `tern_default_labels` is a named `character` vector of available default labels, with each element
#'   named for their corresponding statistic.
#'
#' @export
tern_default_labels <- c(
  fraction = "fraction",
  unique = "Number of patients with at least one event",
  nonunique = "Number of events",
  n = "n",
  count = "count",
  count_fraction = "count_fraction",
  count_fraction_fixed_dp = "count_fraction",
  n_blq = "n_blq",
  sum = "Sum",
  mean = "Mean",
  sd = "SD",
  se = "SE",
  mean_sd = "Mean (SD)",
  mean_se = "Mean (SE)",
  mean_ci = "Mean 95% CI",
  mean_sei = "Mean -/+ 1xSE",
  mean_sdi = "Mean -/+ 1xSD",
  mean_pval = "Mean p-value (H0: mean = 0)",
  median = "Median",
  mad = "Median Absolute Deviation",
  median_ci = "Median 95% CI",
  quantiles = "25% and 75%-ile",
  iqr = "IQR",
  range = "Min - Max",
  min = "Minimum",
  max = "Maximum",
  median_range = "Median (Min - Max)",
  cv = "CV (%)",
  geom_mean = "Geometric Mean",
  geom_mean_ci = "Geometric Mean 95% CI",
  geom_cv = "CV % Geometric Mean",
  pval = "p-value (t-test)", # Default for numeric
  pval_counts = "p-value (chi-squared test)", # Default for counts
  rate = "Adjusted Rate",
  rate_ratio = "Adjusted Rate Ratio"
)

# To deprecate ---------

#' @describeIn default_stats_formats_labels `r lifecycle::badge("deprecated")`
#'   Quick function to retrieve default formats for summary statistics:
#'   [analyze_vars()] and [analyze_vars_in_cols()] principally.
#'
#' @param type (`string`)\cr `"numeric"` or `"counts"`.
#'
#' @return
#' * `summary_formats()` returns a named `vector` of default statistic formats for the given data type.
#'
#' @examples
#' summary_formats()
#' summary_formats(type = "counts", include_pval = TRUE)
#'
#' @export
summary_formats <- function(type = "numeric", include_pval = FALSE) {
  lifecycle::deprecate_warn(
    "0.9.6", "summary_formats()",
    details = 'Use get_formats_from_stats(get_stats("analyze_vars_numeric", add_pval = include_pval)) instead'
  )
  met_grp <- paste0(c("analyze_vars", type), collapse = "_")
  get_formats_from_stats(get_stats(met_grp, add_pval = include_pval))
}

#' @describeIn default_stats_formats_labels `r lifecycle::badge("deprecated")`
#'   Quick function to retrieve default labels for summary statistics.
#'   Returns labels of descriptive statistics which are understood by `rtables`. Similar to `summary_formats`.
#'
#' @param include_pval (`flag`)\cr same as the `add_pval` argument in [get_stats()].
#'
#' @return
#' * `summary_labels` returns a named `vector` of default statistic labels for the given data type.
#'
#' @examples
#' summary_labels()
#' summary_labels(type = "counts", include_pval = TRUE)
#'
#' @export
summary_labels <- function(type = "numeric", include_pval = FALSE) {
  lifecycle::deprecate_warn(
    "0.9.6", "summary_formats()",
    details = 'Use get_labels_from_stats(get_stats("analyze_vars_numeric", add_pval = include_pval)) instead'
  )
  met_grp <- paste0(c("analyze_vars", type), collapse = "_")
  get_labels_from_stats(get_stats(met_grp, add_pval = include_pval))
}

#' Helper functions for tabulating survival duration by subgroup
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Helper functions that tabulate in a data frame statistics such as median survival
#' time and hazard ratio for population subgroups.
#'
#' @inheritParams argument_convention
#' @inheritParams survival_coxph_pairwise
#' @inheritParams survival_duration_subgroups
#' @param arm (`factor`)\cr the treatment group variable.
#'
#' @details Main functionality is to prepare data for use in a layout-creating function.
#'
#' @examples
#' library(dplyr)
#' library(forcats)
#'
#' adtte <- tern_ex_adtte
#'
#' # Save variable labels before data processing steps.
#' adtte_labels <- formatters::var_labels(adtte)
#'
#' adtte_f <- adtte %>%
#'   filter(
#'     PARAMCD == "OS",
#'     ARM %in% c("B: Placebo", "A: Drug X"),
#'     SEX %in% c("M", "F")
#'   ) %>%
#'   mutate(
#'     # Reorder levels of ARM to display reference arm before treatment arm.
#'     ARM = droplevels(fct_relevel(ARM, "B: Placebo")),
#'     SEX = droplevels(SEX),
#'     is_event = CNSR == 0
#'   )
#' labels <- c("ARM" = adtte_labels[["ARM"]], "SEX" = adtte_labels[["SEX"]], "is_event" = "Event Flag")
#' formatters::var_labels(adtte_f)[names(labels)] <- labels
#'
#' @name h_survival_duration_subgroups
NULL

#' @describeIn h_survival_duration_subgroups Helper to prepare a data frame of median survival times by arm.
#'
#' @return
#' * `h_survtime_df()` returns a `data.frame` with columns `arm`, `n`, `n_events`, and `median`.
#'
#' @examples
#' # Extract median survival time for one group.
#' h_survtime_df(
#'   tte = adtte_f$AVAL,
#'   is_event = adtte_f$is_event,
#'   arm = adtte_f$ARM
#' )
#'
#' @export
h_survtime_df <- function(tte, is_event, arm) {
  checkmate::assert_numeric(tte)
  checkmate::assert_logical(is_event, len = length(tte))
  assert_valid_factor(arm, len = length(tte))

  df_tte <- data.frame(
    tte = tte,
    is_event = is_event,
    stringsAsFactors = FALSE
  )

  # Delete NAs
  non_missing_rows <- stats::complete.cases(df_tte)
  df_tte <- df_tte[non_missing_rows, ]
  arm <- arm[non_missing_rows]

  lst_tte <- split(df_tte, arm)
  lst_results <- Map(function(x, arm) {
    if (nrow(x) > 0) {
      s_surv <- s_surv_time(x, .var = "tte", is_event = "is_event")
      median_est <- unname(as.numeric(s_surv$median))
      n_events <- sum(x$is_event)
    } else {
      median_est <- NA
      n_events <- NA
    }

    data.frame(
      arm = arm,
      n = nrow(x),
      n_events = n_events,
      median = median_est,
      stringsAsFactors = FALSE
    )
  }, lst_tte, names(lst_tte))

  df <- do.call(rbind, args = c(lst_results, make.row.names = FALSE))
  df$arm <- factor(df$arm, levels = levels(arm))
  df
}

#' @describeIn h_survival_duration_subgroups Summarizes median survival times by arm and across subgroups
#'    in a data frame. `variables` corresponds to the names of variables found in `data`, passed as a named list and
#'    requires elements `tte`, `is_event`, `arm` and optionally `subgroups`. `groups_lists` optionally specifies
#'    groupings for `subgroups` variables.
#'
#' @return
#' * `h_survtime_subgroups_df()` returns a `data.frame` with columns `arm`, `n`, `n_events`, `median`, `subgroup`,
#'   `var`, `var_label`, and `row_type`.
#'
#' @examples
#' # Extract median survival time for multiple groups.
#' h_survtime_subgroups_df(
#'   variables = list(
#'     tte = "AVAL",
#'     is_event = "is_event",
#'     arm = "ARM",
#'     subgroups = c("SEX", "BMRKR2")
#'   ),
#'   data = adtte_f
#' )
#'
#' # Define groupings for BMRKR2 levels.
#' h_survtime_subgroups_df(
#'   variables = list(
#'     tte = "AVAL",
#'     is_event = "is_event",
#'     arm = "ARM",
#'     subgroups = c("SEX", "BMRKR2")
#'   ),
#'   data = adtte_f,
#'   groups_lists = list(
#'     BMRKR2 = list(
#'       "low" = "LOW",
#'       "low/medium" = c("LOW", "MEDIUM"),
#'       "low/medium/high" = c("LOW", "MEDIUM", "HIGH")
#'     )
#'   )
#' )
#'
#' @export
h_survtime_subgroups_df <- function(variables,
                                    data,
                                    groups_lists = list(),
                                    label_all = "All Patients") {
  checkmate::assert_character(variables$tte)
  checkmate::assert_character(variables$is_event)
  checkmate::assert_character(variables$arm)
  checkmate::assert_character(variables$subgroups, null.ok = TRUE)

  assert_df_with_variables(data, variables)

  checkmate::assert_string(label_all)

  # Add All Patients.
  result_all <- h_survtime_df(data[[variables$tte]], data[[variables$is_event]], data[[variables$arm]])
  result_all$subgroup <- label_all
  result_all$var <- "ALL"
  result_all$var_label <- label_all
  result_all$row_type <- "content"

  # Add Subgroups.
  if (is.null(variables$subgroups)) {
    result_all
  } else {
    l_data <- h_split_by_subgroups(data, variables$subgroups, groups_lists = groups_lists)
    l_result <- lapply(l_data, function(grp) {
      result <- h_survtime_df(grp$df[[variables$tte]], grp$df[[variables$is_event]], grp$df[[variables$arm]])
      result_labels <- grp$df_labels[rep(1, times = nrow(result)), ]
      cbind(result, result_labels)
    })
    result_subgroups <- do.call(rbind, args = c(l_result, make.row.names = FALSE))
    result_subgroups$row_type <- "analysis"
    rbind(
      result_all,
      result_subgroups
    )
  }
}

#' @describeIn h_survival_duration_subgroups Helper to prepare a data frame with estimates of
#'   treatment hazard ratio.
#'
#' @param strata_data (`factor`, `data.frame`, or `NULL`)\cr required if stratified analysis is performed.
#'
#' @return
#' * `h_coxph_df()` returns a `data.frame` with columns `arm`, `n_tot`, `n_tot_events`, `hr`, `lcl`, `ucl`,
#'   `conf_level`, `pval` and `pval_label`.
#'
#' @examples
#' # Extract hazard ratio for one group.
#' h_coxph_df(adtte_f$AVAL, adtte_f$is_event, adtte_f$ARM)
#'
#' # Extract hazard ratio for one group with stratification factor.
#' h_coxph_df(adtte_f$AVAL, adtte_f$is_event, adtte_f$ARM, strata_data = adtte_f$STRATA1)
#'
#' @export
h_coxph_df <- function(tte, is_event, arm, strata_data = NULL, control = control_coxph()) {
  checkmate::assert_numeric(tte)
  checkmate::assert_logical(is_event, len = length(tte))
  assert_valid_factor(arm, n.levels = 2, len = length(tte))

  df_tte <- data.frame(tte = tte, is_event = is_event)
  strata_vars <- NULL

  if (!is.null(strata_data)) {
    if (is.data.frame(strata_data)) {
      strata_vars <- names(strata_data)
      checkmate::assert_data_frame(strata_data, nrows = nrow(df_tte))
      assert_df_with_factors(strata_data, as.list(stats::setNames(strata_vars, strata_vars)))
    } else {
      assert_valid_factor(strata_data, len = nrow(df_tte))
      strata_vars <- "strata_data"
    }
    df_tte[strata_vars] <- strata_data
  }

  l_df <- split(df_tte, arm)

  if (nrow(l_df[[1]]) > 0 && nrow(l_df[[2]]) > 0) {
    # Hazard ratio and CI.
    result <- s_coxph_pairwise(
      df = l_df[[2]],
      .ref_group = l_df[[1]],
      .in_ref_col = FALSE,
      .var = "tte",
      is_event = "is_event",
      strata = strata_vars,
      control = control
    )

    df <- data.frame(
      # Dummy column needed downstream to create a nested header.
      arm = " ",
      n_tot = unname(as.numeric(result$n_tot)),
      n_tot_events = unname(as.numeric(result$n_tot_events)),
      hr = unname(as.numeric(result$hr)),
      lcl = unname(result$hr_ci[1]),
      ucl = unname(result$hr_ci[2]),
      conf_level = control[["conf_level"]],
      pval = as.numeric(result$pvalue),
      pval_label = obj_label(result$pvalue),
      stringsAsFactors = FALSE
    )
  } else if (
    (nrow(l_df[[1]]) == 0 && nrow(l_df[[2]]) > 0) ||
      (nrow(l_df[[1]]) > 0 && nrow(l_df[[2]]) == 0)
  ) {
    df_tte_complete <- df_tte[stats::complete.cases(df_tte), ]
    df <- data.frame(
      # Dummy column needed downstream to create a nested header.
      arm = " ",
      n_tot = nrow(df_tte_complete),
      n_tot_events = sum(df_tte_complete$is_event),
      hr = NA,
      lcl = NA,
      ucl = NA,
      conf_level = control[["conf_level"]],
      pval = NA,
      pval_label = NA,
      stringsAsFactors = FALSE
    )
  } else {
    df <- data.frame(
      # Dummy column needed downstream to create a nested header.
      arm = " ",
      n_tot = 0L,
      n_tot_events = 0L,
      hr = NA,
      lcl = NA,
      ucl = NA,
      conf_level = control[["conf_level"]],
      pval = NA,
      pval_label = NA,
      stringsAsFactors = FALSE
    )
  }

  df
}

#' @describeIn h_survival_duration_subgroups Summarizes estimates of the treatment hazard ratio
#'   across subgroups in a data frame. `variables` corresponds to the names of variables found in
#'   `data`, passed as a named list and requires elements `tte`, `is_event`, `arm` and
#'   optionally `subgroups` and `strata`. `groups_lists` optionally specifies
#'   groupings for `subgroups` variables.
#'
#' @return
#' * `h_coxph_subgroups_df()` returns a `data.frame` with columns `arm`, `n_tot`, `n_tot_events`, `hr`,
#'   `lcl`, `ucl`, `conf_level`, `pval`, `pval_label`, `subgroup`, `var`, `var_label`, and `row_type`.
#'
#' @examples
#' # Extract hazard ratio for multiple groups.
#' h_coxph_subgroups_df(
#'   variables = list(
#'     tte = "AVAL",
#'     is_event = "is_event",
#'     arm = "ARM",
#'     subgroups = c("SEX", "BMRKR2")
#'   ),
#'   data = adtte_f
#' )
#'
#' # Define groupings of BMRKR2 levels.
#' h_coxph_subgroups_df(
#'   variables = list(
#'     tte = "AVAL",
#'     is_event = "is_event",
#'     arm = "ARM",
#'     subgroups = c("SEX", "BMRKR2")
#'   ),
#'   data = adtte_f,
#'   groups_lists = list(
#'     BMRKR2 = list(
#'       "low" = "LOW",
#'       "low/medium" = c("LOW", "MEDIUM"),
#'       "low/medium/high" = c("LOW", "MEDIUM", "HIGH")
#'     )
#'   )
#' )
#'
#' # Extract hazard ratio for multiple groups with stratification factors.
#' h_coxph_subgroups_df(
#'   variables = list(
#'     tte = "AVAL",
#'     is_event = "is_event",
#'     arm = "ARM",
#'     subgroups = c("SEX", "BMRKR2"),
#'     strata = c("STRATA1", "STRATA2")
#'   ),
#'   data = adtte_f
#' )
#'
#' @export
h_coxph_subgroups_df <- function(variables,
                                 data,
                                 groups_lists = list(),
                                 control = control_coxph(),
                                 label_all = "All Patients") {
  if ("strat" %in% names(variables)) {
    warning(
      "Warning: the `strat` element name of the `variables` list argument to `h_coxph_subgroups_df() ",
      "was deprecated in tern 0.9.4.\n  ",
      "Please use the name `strata` instead of `strat` in the `variables` argument."
    )
    variables[["strata"]] <- variables[["strat"]]
  }

  checkmate::assert_character(variables$tte)
  checkmate::assert_character(variables$is_event)
  checkmate::assert_character(variables$arm)
  checkmate::assert_character(variables$subgroups, null.ok = TRUE)
  checkmate::assert_character(variables$strata, null.ok = TRUE)
  assert_df_with_factors(data, list(val = variables$arm), min.levels = 2, max.levels = 2)
  assert_df_with_variables(data, variables)
  checkmate::assert_string(label_all)

  # Add All Patients.
  result_all <- h_coxph_df(
    tte = data[[variables$tte]],
    is_event = data[[variables$is_event]],
    arm = data[[variables$arm]],
    strata_data = if (is.null(variables$strata)) NULL else data[variables$strata],
    control = control
  )
  result_all$subgroup <- label_all
  result_all$var <- "ALL"
  result_all$var_label <- label_all
  result_all$row_type <- "content"

  # Add Subgroups.
  if (is.null(variables$subgroups)) {
    result_all
  } else {
    l_data <- h_split_by_subgroups(data, variables$subgroups, groups_lists = groups_lists)

    l_result <- lapply(l_data, function(grp) {
      result <- h_coxph_df(
        tte = grp$df[[variables$tte]],
        is_event = grp$df[[variables$is_event]],
        arm = grp$df[[variables$arm]],
        strata_data = if (is.null(variables$strata)) NULL else grp$df[variables$strata],
        control = control
      )
      result_labels <- grp$df_labels[rep(1, times = nrow(result)), ]
      cbind(result, result_labels)
    })

    result_subgroups <- do.call(rbind, args = c(l_result, make.row.names = FALSE))
    result_subgroups$row_type <- "analysis"

    rbind(
      result_all,
      result_subgroups
    )
  }
}

#' Split data frame by subgroups
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Split a data frame into a non-nested list of subsets.
#'
#' @inheritParams argument_convention
#' @inheritParams survival_duration_subgroups
#' @param data (`data.frame`)\cr dataset to split.
#' @param subgroups (`character`)\cr names of factor variables from `data` used to create subsets.
#'   Unused levels not present in `data` are dropped. Note that the order in this vector
#'   determines the order in the downstream table.
#'
#' @return A list with subset data (`df`) and metadata about the subset (`df_labels`).
#'
#' @details Main functionality is to prepare data for use in forest plot layouts.
#'
#' @examples
#' df <- data.frame(
#'   x = c(1:5),
#'   y = factor(c("A", "B", "A", "B", "A"), levels = c("A", "B", "C")),
#'   z = factor(c("C", "C", "D", "D", "D"), levels = c("D", "C"))
#' )
#' formatters::var_labels(df) <- paste("label for", names(df))
#'
#' h_split_by_subgroups(
#'   data = df,
#'   subgroups = c("y", "z")
#' )
#'
#' h_split_by_subgroups(
#'   data = df,
#'   subgroups = c("y", "z"),
#'   groups_lists = list(
#'     y = list("AB" = c("A", "B"), "C" = "C")
#'   )
#' )
#'
#' @export
h_split_by_subgroups <- function(data,
                                 subgroups,
                                 groups_lists = list()) {
  checkmate::assert_character(subgroups, min.len = 1, any.missing = FALSE)
  checkmate::assert_list(groups_lists, names = "named")
  checkmate::assert_subset(names(groups_lists), subgroups)
  assert_df_with_factors(data, as.list(stats::setNames(subgroups, subgroups)))

  data_labels <- unname(formatters::var_labels(data))
  df_subgroups <- data[, subgroups, drop = FALSE]
  subgroup_labels <- formatters::var_labels(df_subgroups, fill = TRUE)

  l_labels <- Map(function(grp_i, name_i) {
    existing_levels <- levels(droplevels(grp_i))
    grp_levels <- if (name_i %in% names(groups_lists)) {
      # For this variable groupings are defined. We check which groups are contained in the data.
      group_list_i <- groups_lists[[name_i]]
      group_has_levels <- vapply(group_list_i, function(lvls) any(lvls %in% existing_levels), TRUE)
      names(which(group_has_levels))
    } else {
      existing_levels
    }
    df_labels <- data.frame(
      subgroup = grp_levels,
      var = name_i,
      var_label = unname(subgroup_labels[name_i]),
      stringsAsFactors = FALSE # Rationale is that subgroups may not be unique.
    )
  }, df_subgroups, names(df_subgroups))

  # Create a data frame with one row per subgroup.
  df_labels <- do.call(rbind, args = c(l_labels, make.row.names = FALSE))
  row_label <- paste0(df_labels$var, ".", df_labels$subgroup)
  row_split_var <- factor(row_label, levels = row_label)

  # Create a list of data subsets.
  lapply(split(df_labels, row_split_var), function(row_i) {
    which_row <- if (row_i$var %in% names(groups_lists)) {
      data[[row_i$var]] %in% groups_lists[[row_i$var]][[row_i$subgroup]]
    } else {
      data[[row_i$var]] == row_i$subgroup
    }
    df <- data[which_row, ]
    rownames(df) <- NULL
    formatters::var_labels(df) <- data_labels

    list(
      df = df,
      df_labels = data.frame(row_i, row.names = NULL)
    )
  })
}

#' Survival time analysis
#'
#' @description `r lifecycle::badge("stable")`
#'
#' The analyze function [surv_time()] creates a layout element to analyze survival time by calculating survival time
#' median, median confidence interval, quantiles, and range (for all, censored, or event patients). The primary
#' analysis variable `vars` is the time variable and the secondary analysis variable `is_event` indicates whether or
#' not an event has occurred.
#'
#' @inheritParams argument_convention
#' @param control (`list`)\cr parameters for comparison details, specified by using the helper function
#'   [control_surv_time()]. Some possible parameter options are:
#'   * `conf_level` (`proportion`)\cr confidence level of the interval for survival time.
#'   * `conf_type` (`string`)\cr confidence interval type. Options are "plain" (default), "log", or "log-log",
#'     see more in [survival::survfit()]. Note option "none" is not supported.
#'   * `quantiles` (`numeric`)\cr vector of length two to specify the quantiles of survival time.
#' @param ref_fn_censor (`flag`)\cr whether referential footnotes indicating censored observations should be printed
#'   when the `range` statistic is included.
#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("surv_time")`
#'   to see available statistics for this function.
#' @param .indent_mods (named `integer`)\cr indent modifiers for the labels. Each element of the vector
#'   should be a name-value pair with name corresponding to a statistic specified in `.stats` and value the indentation
#'   for that statistic's row label.
#'
#' @examples
#' library(dplyr)
#'
#' adtte_f <- tern_ex_adtte %>%
#'   filter(PARAMCD == "OS") %>%
#'   mutate(
#'     AVAL = day2month(AVAL),
#'     is_event = CNSR == 0
#'   )
#' df <- adtte_f %>% filter(ARMCD == "ARM A")
#'
#' @name survival_time
#' @order 1
NULL

#' @describeIn survival_time Statistics function which analyzes survival times.
#'
#' @return
#' * `s_surv_time()` returns the statistics:
#'   * `median`: Median survival time.
#'   * `median_ci`: Confidence interval for median time.
#'   * `quantiles`: Survival time for two specified quantiles.
#'   * `range_censor`: Survival time range for censored observations.
#'   * `range_event`: Survival time range for observations with events.
#'   * `range`: Survival time range for all observations.
#'
#' @keywords internal
s_surv_time <- function(df,
                        .var,
                        is_event,
                        control = control_surv_time()) {
  checkmate::assert_string(.var)
  assert_df_with_variables(df, list(tte = .var, is_event = is_event))
  checkmate::assert_numeric(df[[.var]], min.len = 1, any.missing = FALSE)
  checkmate::assert_logical(df[[is_event]], min.len = 1, any.missing = FALSE)

  conf_type <- control$conf_type
  conf_level <- control$conf_level
  quantiles <- control$quantiles

  formula <- stats::as.formula(paste0("survival::Surv(", .var, ", ", is_event, ") ~ 1"))
  srv_fit <- survival::survfit(
    formula = formula,
    data = df,
    conf.int = conf_level,
    conf.type = conf_type
  )
  srv_tab <- summary(srv_fit, extend = TRUE)$table
  srv_qt_tab <- stats::quantile(srv_fit, probs = quantiles)$quantile
  range_censor <- range_noinf(df[[.var]][!df[[is_event]]], na.rm = TRUE)
  range_event <- range_noinf(df[[.var]][df[[is_event]]], na.rm = TRUE)
  range <- range_noinf(df[[.var]], na.rm = TRUE)
  list(
    median = formatters::with_label(unname(srv_tab["median"]), "Median"),
    median_ci = formatters::with_label(
      unname(srv_tab[paste0(srv_fit$conf.int, c("LCL", "UCL"))]), f_conf_level(conf_level)
    ),
    quantiles = formatters::with_label(
      unname(srv_qt_tab), paste0(quantiles[1] * 100, "% and ", quantiles[2] * 100, "%-ile")
    ),
    range_censor = formatters::with_label(range_censor, "Range (censored)"),
    range_event = formatters::with_label(range_event, "Range (event)"),
    range = formatters::with_label(range, "Range")
  )
}

#' @describeIn survival_time Formatted analysis function which is used as `afun` in `surv_time()`.
#'
#' @return
#' * `a_surv_time()` returns the corresponding list with formatted [rtables::CellValue()].
#'
#' @examples
#' a_surv_time(
#'   df,
#'   .df_row = df,
#'   .var = "AVAL",
#'   is_event = "is_event"
#' )
#'
#' @export
a_surv_time <- function(df,
                        labelstr = "",
                        .var = NULL,
                        .df_row = NULL,
                        is_event,
                        control = control_surv_time(),
                        ref_fn_censor = TRUE,
                        .stats = NULL,
                        .formats = NULL,
                        .labels = NULL,
                        .indent_mods = NULL,
                        na_str = default_na_str()) {
  x_stats <- s_surv_time(
    df = df, .var = .var, is_event = is_event, control = control
  )
  rng_censor_lwr <- x_stats[["range_censor"]][1]
  rng_censor_upr <- x_stats[["range_censor"]][2]

  # Use method-specific defaults
  fmts <- c(median_ci = "(xx.x, xx.x)", quantiles = "xx.x, xx.x", range = "xx.x to xx.x")
  lbls <- c(median_ci = "95% CI", range = "Range", range_censor = "Range (censored)", range_event = "Range (event)")
  lbls_custom <- .labels
  .formats <- c(.formats, fmts[setdiff(names(fmts), names(.formats))])
  .labels <- c(.labels, lbls[setdiff(names(lbls), names(lbls_custom))])

  # Fill in with formatting defaults if needed
  .stats <- get_stats("surv_time", stats_in = .stats)
  .formats <- get_formats_from_stats(.stats, .formats)
  .labels <- get_labels_from_stats(.stats, .labels) %>% labels_use_control(control, lbls_custom)
  .indent_mods <- get_indents_from_stats(.stats, .indent_mods)

  x_stats <- x_stats[.stats]

  # Auto format handling
  .formats <- apply_auto_formatting(.formats, x_stats, .df_row, .var)

  cell_fns <- setNames(vector("list", length = length(x_stats)), .labels)
  if ("range" %in% names(x_stats) && ref_fn_censor) {
    if (identical(x_stats[["range"]][1], rng_censor_lwr) && identical(x_stats[["range"]][2], rng_censor_upr)) {
      cell_fns[[.labels[["range"]]]] <- "Censored observations: range minimum & maximum"
    } else if (identical(x_stats[["range"]][1], rng_censor_lwr)) {
      cell_fns[[.labels[["range"]]]] <- "Censored observation: range minimum"
    } else if (identical(x_stats[["range"]][2], rng_censor_upr)) {
      cell_fns[[.labels[["range"]]]] <- "Censored observation: range maximum"
    }
  }

  in_rows(
    .list = x_stats,
    .formats = .formats,
    .names = .labels,
    .labels = .labels,
    .indent_mods = .indent_mods,
    .format_na_strs = na_str,
    .cell_footnotes = cell_fns
  )
}

#' @describeIn survival_time Layout-creating function which can take statistics function arguments
#'   and additional format arguments. This function is a wrapper for [rtables::analyze()].
#'
#' @return
#' * `surv_time()` returns a layout object suitable for passing to further layouting functions,
#'   or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
#'   the statistics from `s_surv_time()` to the table layout.
#'
#' @examples
#' basic_table() %>%
#'   split_cols_by(var = "ARMCD") %>%
#'   add_colcounts() %>%
#'   surv_time(
#'     vars = "AVAL",
#'     var_labels = "Survival Time (Months)",
#'     is_event = "is_event",
#'     control = control_surv_time(conf_level = 0.9, conf_type = "log-log")
#'   ) %>%
#'   build_table(df = adtte_f)
#'
#' @export
#' @order 2
surv_time <- function(lyt,
                      vars,
                      is_event,
                      control = control_surv_time(),
                      ref_fn_censor = TRUE,
                      na_str = default_na_str(),
                      nested = TRUE,
                      ...,
                      var_labels = "Time to Event",
                      show_labels = "visible",
                      table_names = vars,
                      .stats = c("median", "median_ci", "quantiles", "range"),
                      .formats = NULL,
                      .labels = NULL,
                      .indent_mods = c(median_ci = 1L)) {
  extra_args <- list(
    .stats = .stats, .formats = .formats, .labels = .labels, .indent_mods = .indent_mods, na_str = na_str,
    is_event = is_event, control = control, ref_fn_censor = ref_fn_censor, ...
  )

  analyze(
    lyt = lyt,
    vars = vars,
    afun = a_surv_time,
    var_labels = var_labels,
    show_labels = show_labels,
    table_names = table_names,
    na_str = na_str,
    nested = nested,
    extra_args = extra_args
  )
}

#' Line plot with optional table
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Line plot with the optional table.
#'
#' @inheritParams argument_convention
#' @param alt_counts_df (`data.frame` or `NULL`)\cr data set that will be used (only)
#'   to counts objects in groups for stratification.
#' @param variables (named `character`) vector of variable names in `df` which should include:
#'   * `x` (`string`)\cr name of x-axis variable.
#'   * `y` (`string`)\cr name of y-axis variable.
#'   * `group_var` (`string` or `NULL`)\cr name of grouping variable (or strata), i.e. treatment arm.
#'     Can be `NA` to indicate lack of groups.
#'   * `subject_var` (`string` or `NULL`)\cr name of subject variable. Only applies if `group_var` is
#'      not NULL.
#'   * `paramcd` (`string` or `NA`)\cr name of the variable for parameter's code. Used for y-axis label and plot's
#'     subtitle. Can be `NA` if `paramcd` is not to be added to the y-axis label or subtitle.
#'   * `y_unit` (`string` or `NA`)\cr name of variable with units of `y`. Used for y-axis label and plot's subtitle.
#'     Can be `NA` if y unit is not to be added to the y-axis label or subtitle.
#'   * `facet_var` (`string` or `NA`)\cr name of the secondary grouping variable used for plot faceting, i.e. treatment
#'     arm. Can be `NA` to indicate lack of groups.
#' @param mid (`character` or `NULL`)\cr names of the statistics that will be plotted as midpoints.
#'   All the statistics indicated in `mid` variable must be present in the object returned by `sfun`,
#'   and be of a `double` or `numeric` type vector of length one.
#' @param interval (`character` or `NULL`)\cr names of the statistics that will be plotted as intervals.
#'   All the statistics indicated in `interval` variable must be present in the object returned by `sfun`,
#'   and be of a `double` or `numeric` type vector of length two. Set `interval = NULL` if intervals should not be
#'   added to the plot.
#' @param whiskers (`character`)\cr names of the interval whiskers that will be plotted. Names must match names
#'   of the list element `interval` that will be returned by `sfun` (e.g. `mean_ci_lwr` element of
#'   `sfun(x)[["mean_ci"]]`). It is possible to specify one whisker only, or to suppress all whiskers by setting
#'   `interval = NULL`.
#' @param table (`character` or `NULL`)\cr names of the statistics that will be displayed in the table below the plot.
#'   All the statistics indicated in `table` variable must be present in the object returned by `sfun`.
#' @param sfun (`function`)\cr the function to compute the values of required statistics. It must return a named `list`
#'   with atomic vectors. The names of the `list` elements refer to the names of the statistics and are used by `mid`,
#'   `interval`, `table`. It must be able to accept as input a vector with data for which statistics are computed.
#' @param ... optional arguments to `sfun`.
#' @param mid_type (`string`)\cr controls the type of the `mid` plot, it can be point (`"p"`), line (`"l"`),
#'   or point and line (`"pl"`).
#' @param mid_point_size (`numeric(1)`)\cr font size of the `mid` plot points.
#' @param position (`character` or `call`)\cr geom element position adjustment, either as a string, or the result of
#'   a call to a position adjustment function.
#' @param legend_title (`string`)\cr legend title.
#' @param legend_position (`string`)\cr the position of the plot legend (`"none"`, `"left"`, `"right"`, `"bottom"`,
#'   `"top"`, or a two-element numeric vector).
#' @param ggtheme (`theme`)\cr a graphical theme as provided by `ggplot2` to control styling of the plot.
#' @param xticks (`numeric` or `NULL`)\cr numeric vector of tick positions or a single number with spacing
#'   between ticks on the x-axis, for use when `variables$x` is numeric. If `NULL` (default), [labeling::extended()] is
#'   used to determine optimal tick positions on the x-axis. If `variables$x` is not numeric, this argument is ignored.
#' @param x_lab (`string` or `NULL`)\cr x-axis label. If `NULL` then no label will be added.
#' @param y_lab (`string` or `NULL`)\cr y-axis label. If `NULL` then no label will be added.
#' @param y_lab_add_paramcd (`flag`)\cr whether `paramcd`, i.e. `unique(df[[variables["paramcd"]]])` should be added
#'   to the y-axis label (`y_lab`).
#' @param y_lab_add_unit (`flag`)\cr whether y-axis unit, i.e. `unique(df[[variables["y_unit"]]])` should be added
#'   to the y-axis label (`y_lab`).
#' @param title (`string`)\cr plot title.
#' @param subtitle (`string`)\cr plot subtitle.
#' @param subtitle_add_paramcd (`flag`)\cr whether `paramcd`, i.e. `unique(df[[variables["paramcd"]]])` should be
#'   added to the plot's subtitle (`subtitle`).
#' @param subtitle_add_unit (`flag`)\cr whether the y-axis unit, i.e. `unique(df[[variables["y_unit"]]])` should be
#'   added to the plot's subtitle (`subtitle`).
#' @param caption (`string`)\cr optional caption below the plot.
#' @param table_format (named `character` or `NULL`)\cr format patterns for descriptive statistics used in the
#'   (optional) table appended to the plot. It is passed directly to the `h_format_row` function through the `format`
#'   parameter. Names of `table_format` must match the names of statistics returned by `sfun` function.
#' @param table_labels (named `character` or `NULL`)\cr labels for descriptive statistics used in the (optional) table
#'   appended to the plot. Names of `table_labels` must match the names of statistics returned by `sfun` function.
#' @param table_font_size (`numeric(1)`)\cr font size of the text in the table.
#' @param newpage `r lifecycle::badge("deprecated")` not used.
#' @param col (`character`)\cr color(s). See `?ggplot2::aes_colour_fill_alpha` for example values.
#' @param linetype (`character`)\cr line type(s). See `?ggplot2::aes_linetype_size_shape` for example values.
#' @param errorbar_width (`numeric(1)`)\cr width of the error bars.
#'
#' @return A `ggplot` line plot (and statistics table if applicable).
#'
#' @examples
#' library(nestcolor)
#'
#' adsl <- tern_ex_adsl
#' adlb <- tern_ex_adlb %>% dplyr::filter(ANL01FL == "Y", PARAMCD == "ALT", AVISIT != "SCREENING")
#' adlb$AVISIT <- droplevels(adlb$AVISIT)
#' adlb <- dplyr::mutate(adlb, AVISIT = forcats::fct_reorder(AVISIT, AVISITN, min))
#'
#' # Mean with CI
#' g_lineplot(adlb, adsl, subtitle = "Laboratory Test:")
#'
#' # Mean with CI, no stratification with group_var
#' g_lineplot(adlb, variables = control_lineplot_vars(group_var = NA))
#'
#' # Mean, upper whisker of CI, no group_var(strata) counts N
#' g_lineplot(
#'   adlb,
#'   whiskers = "mean_ci_upr",
#'   title = "Plot of Mean and Upper 95% Confidence Limit by Visit"
#' )
#'
#' # Median with CI
#' g_lineplot(
#'   adlb,
#'   adsl,
#'   mid = "median",
#'   interval = "median_ci",
#'   whiskers = c("median_ci_lwr", "median_ci_upr"),
#'   title = "Plot of Median and 95% Confidence Limits by Visit"
#' )
#'
#' # Mean, +/- SD
#' g_lineplot(adlb, adsl,
#'   interval = "mean_sdi",
#'   whiskers = c("mean_sdi_lwr", "mean_sdi_upr"),
#'   title = "Plot of Median +/- SD by Visit"
#' )
#'
#' # Mean with CI plot with stats table
#' g_lineplot(adlb, adsl, table = c("n", "mean", "mean_ci"))
#'
#' # Mean with CI, table and customized confidence level
#' g_lineplot(
#'   adlb,
#'   adsl,
#'   table = c("n", "mean", "mean_ci"),
#'   control = control_analyze_vars(conf_level = 0.80),
#'   title = "Plot of Mean and 80% Confidence Limits by Visit"
#' )
#'
#' # Mean with CI, table, filtered data
#' adlb_f <- dplyr::filter(adlb, ARMCD != "ARM A" | AVISIT == "BASELINE")
#' g_lineplot(adlb_f, table = c("n", "mean"))
#'
#' @export
g_lineplot <- function(df,
                       alt_counts_df = NULL,
                       variables = control_lineplot_vars(),
                       mid = "mean",
                       interval = "mean_ci",
                       whiskers = c("mean_ci_lwr", "mean_ci_upr"),
                       table = NULL,
                       sfun = s_summary,
                       ...,
                       mid_type = "pl",
                       mid_point_size = 2,
                       position = ggplot2::position_dodge(width = 0.4),
                       legend_title = NULL,
                       legend_position = "bottom",
                       ggtheme = nestcolor::theme_nest(),
                       xticks = NULL,
                       xlim = NULL,
                       ylim = NULL,
                       x_lab = obj_label(df[[variables[["x"]]]]),
                       y_lab = NULL,
                       y_lab_add_paramcd = TRUE,
                       y_lab_add_unit = TRUE,
                       title = "Plot of Mean and 95% Confidence Limits by Visit",
                       subtitle = "",
                       subtitle_add_paramcd = TRUE,
                       subtitle_add_unit = TRUE,
                       caption = NULL,
                       table_format = NULL,
                       table_labels = NULL,
                       table_font_size = 3,
                       errorbar_width = 0.45,
                       newpage = lifecycle::deprecated(),
                       col = NULL,
                       linetype = NULL) {
  checkmate::assert_character(variables, any.missing = TRUE)
  checkmate::assert_character(mid, null.ok = TRUE)
  checkmate::assert_character(interval, null.ok = TRUE)
  checkmate::assert_character(col, null.ok = TRUE)
  checkmate::assert_character(linetype, null.ok = TRUE)
  checkmate::assert_numeric(xticks, null.ok = TRUE)
  checkmate::assert_numeric(xlim, finite = TRUE, any.missing = FALSE, len = 2, sorted = TRUE, null.ok = TRUE)
  checkmate::assert_numeric(ylim, finite = TRUE, any.missing = FALSE, len = 2, sorted = TRUE, null.ok = TRUE)
  checkmate::assert_number(errorbar_width, lower = 0)
  checkmate::assert_string(title, null.ok = TRUE)
  checkmate::assert_string(subtitle, null.ok = TRUE)

  if (!is.null(table)) {
    table_format <- get_formats_from_stats(table)
    table_labels <- get_labels_from_stats(table)
  }

  extra_args <- list(...)
  if ("control" %in% names(extra_args)) {
    if (!is.null(table) && all(table_labels == get_labels_from_stats(table))) {
      table_labels <- table_labels %>% labels_use_control(extra_args[["control"]])
    }
  }

  if (is.character(interval)) {
    checkmate::assert_vector(whiskers, min.len = 0, max.len = 2)
  }

  if (length(whiskers) == 1) {
    checkmate::assert_character(mid)
  }

  if (is.character(mid)) {
    checkmate::assert_scalar(mid_type)
    checkmate::assert_subset(mid_type, c("pl", "p", "l"))
  }

  x <- variables[["x"]]
  y <- variables[["y"]]
  paramcd <- variables["paramcd"] # NA if paramcd == NA or it is not in variables
  y_unit <- variables["y_unit"] # NA if y_unit == NA or it is not in variables
  if (is.na(variables["group_var"])) {
    group_var <- NULL # NULL if group_var == NA or it is not in variables
  } else {
    group_var <- variables[["group_var"]]
    subject_var <- variables[["subject_var"]]
  }
  if (is.na(variables["facet_var"])) {
    facet_var <- NULL # NULL if facet_var == NA or it is not in variables
  } else {
    facet_var <- variables[["facet_var"]]
  }
  checkmate::assert_flag(y_lab_add_paramcd, null.ok = TRUE)
  checkmate::assert_flag(subtitle_add_paramcd, null.ok = TRUE)
  if ((!is.null(y_lab) && y_lab_add_paramcd) || (!is.null(subtitle) && subtitle_add_paramcd)) {
    checkmate::assert_false(is.na(paramcd))
    checkmate::assert_scalar(unique(df[[paramcd]]))
  }

  checkmate::assert_flag(y_lab_add_unit, null.ok = TRUE)
  checkmate::assert_flag(subtitle_add_unit, null.ok = TRUE)
  if ((!is.null(y_lab) && y_lab_add_unit) || (!is.null(subtitle) && subtitle_add_unit)) {
    checkmate::assert_false(is.na(y_unit))
    checkmate::assert_scalar(unique(df[[y_unit]]))
  }

  if (!is.null(group_var) && !is.null(alt_counts_df)) {
    checkmate::assert_set_equal(unique(alt_counts_df[[group_var]]), unique(df[[group_var]]))
  }

  ####################################### |
  # ---- Compute required statistics ----
  ####################################### |
  # Remove unused levels for x-axis
  if (is.factor(df[[x]])) {
    df[[x]] <- droplevels(df[[x]])
  }

  if (!is.null(facet_var) && !is.null(group_var)) {
    df_grp <- tidyr::expand(df, .data[[facet_var]], .data[[group_var]], .data[[x]]) # expand based on levels of factors
  } else if (!is.null(group_var)) {
    df_grp <- tidyr::expand(df, .data[[group_var]], .data[[x]]) # expand based on levels of factors
  } else {
    df_grp <- tidyr::expand(df, NULL, .data[[x]])
  }

  df_grp <- df_grp %>%
    dplyr::full_join(y = df[, c(facet_var, group_var, x, y)], by = c(facet_var, group_var, x), multiple = "all") %>%
    dplyr::group_by_at(c(facet_var, group_var, x))

  df_stats <- df_grp %>%
    dplyr::summarise(
      data.frame(t(do.call(c, unname(sfun(.data[[y]])[c(mid, interval)])))),
      .groups = "drop"
    )

  df_stats <- df_stats[!is.na(df_stats[[mid]]), ]

  # add number of objects N in group_var (strata)
  if (!is.null(group_var) && !is.null(alt_counts_df)) {
    strata_N <- paste0(group_var, "_N") # nolint

    df_N <- stats::aggregate(eval(parse(text = subject_var)) ~ eval(parse(text = group_var)), data = alt_counts_df, FUN = function(x) length(unique(x))) # nolint
    colnames(df_N) <- c(group_var, "N") # nolint
    df_N[[strata_N]] <- paste0(df_N[[group_var]], " (N = ", df_N$N, ")") # nolint

    # keep strata factor levels
    matches <- sapply(unique(df_N[[group_var]]), function(x) {
      regex_pattern <- gsub("([][(){}^$.|*+?\\\\])", "\\\\\\1", x)
      unique(df_N[[paste0(group_var, "_N")]])[grepl(
        paste0("^", regex_pattern),
        unique(df_N[[paste0(group_var, "_N")]])
      )]
    })
    df_N[[paste0(group_var, "_N")]] <- factor(df_N[[group_var]]) # nolint
    levels(df_N[[paste0(group_var, "_N")]]) <- unlist(matches) # nolint

    # strata_N should not be in colnames(df_stats)
    checkmate::assert_disjunct(strata_N, colnames(df_stats))

    df_stats <- merge(x = df_stats, y = df_N[, c(group_var, strata_N)], by = group_var)
  } else if (!is.null(group_var)) {
    strata_N <- group_var # nolint
  } else {
    strata_N <- NULL # nolint
  }

  ############################################### |
  # ---- Prepare certain plot's properties. ----
  ############################################### |
  # legend title
  if (is.null(legend_title) && !is.null(group_var) && legend_position != "none") {
    legend_title <- attr(df[[group_var]], "label")
  }

  # y label
  if (!is.null(y_lab)) {
    if (y_lab_add_paramcd) {
      y_lab <- paste(y_lab, unique(df[[paramcd]]))
    }

    if (y_lab_add_unit) {
      y_lab <- paste0(y_lab, " (", unique(df[[y_unit]]), ")")
    }

    y_lab <- trimws(y_lab)
  }

  # subtitle
  if (!is.null(subtitle)) {
    if (subtitle_add_paramcd) {
      subtitle <- paste(subtitle, unique(df[[paramcd]]))
    }

    if (subtitle_add_unit) {
      subtitle <- paste0(subtitle, " (", unique(df[[y_unit]]), ")")
    }

    subtitle <- trimws(subtitle)
  }

  ############################### |
  # ---- Build plot object. ----
  ############################### |
  p <- ggplot2::ggplot(
    data = df_stats,
    mapping = ggplot2::aes(
      x = .data[[x]], y = .data[[mid]],
      color = if (is.null(strata_N)) NULL else .data[[strata_N]],
      shape = if (is.null(strata_N)) NULL else .data[[strata_N]],
      lty = if (is.null(strata_N)) NULL else .data[[strata_N]],
      group = if (is.null(strata_N)) NULL else .data[[strata_N]]
    )
  )

  if (!is.null(group_var) && nlevels(df_stats[[strata_N]]) > 6) {
    p <- p +
      scale_shape_manual(values = seq(15, 15 + nlevels(df_stats[[strata_N]])))
  }

  if (!is.null(mid)) {
    # points
    if (grepl("p", mid_type, fixed = TRUE)) {
      p <- p + ggplot2::geom_point(position = position, size = mid_point_size, na.rm = TRUE)
    }

    # lines - plotted only if there is a strata grouping (group_var)
    if (grepl("l", mid_type, fixed = TRUE) && !is.null(strata_N)) { # nolint
      p <- p + ggplot2::geom_line(position = position, na.rm = TRUE)
    }
  }

  # interval
  if (!is.null(interval)) {
    p <- p +
      ggplot2::geom_errorbar(
        ggplot2::aes(ymin = .data[[whiskers[1]]], ymax = .data[[whiskers[max(1, length(whiskers))]]]),
        width = errorbar_width,
        position = position
      )

    if (length(whiskers) == 1) { # lwr or upr only; mid is then required
      # workaround as geom_errorbar does not provide single-direction whiskers
      p <- p +
        ggplot2::geom_linerange(
          data = df_stats[!is.na(df_stats[[whiskers]]), ], # as na.rm =TRUE does not suppress warnings
          ggplot2::aes(ymin = .data[[mid]], ymax = .data[[whiskers]]),
          position = position,
          na.rm = TRUE,
          show.legend = FALSE
        )
    }
  }

  if (is.numeric(df_stats[[x]])) {
    if (length(xticks) == 1) xticks <- seq(from = min(df_stats[[x]]), to = max(df_stats[[x]]), by = xticks)
    p <- p + ggplot2::scale_x_continuous(breaks = if (!is.null(xticks)) xticks else waiver(), limits = xlim)
  }

  p <- p +
    ggplot2::scale_y_continuous(labels = scales::comma, limits = ylim) +
    ggplot2::labs(
      title = title,
      subtitle = subtitle,
      caption = caption,
      color = legend_title,
      lty = legend_title,
      shape = legend_title,
      x = x_lab,
      y = y_lab
    )

  if (!is.null(col)) {
    p <- p +
      ggplot2::scale_color_manual(values = col)
  }
  if (!is.null(linetype)) {
    p <- p +
      ggplot2::scale_linetype_manual(values = linetype)
  }

  if (!is.null(facet_var)) {
    p <- p +
      facet_grid(cols = vars(df_stats[[facet_var]]))
  }

  if (!is.null(ggtheme)) {
    p <- p + ggtheme
  } else {
    p <- p +
      ggplot2::theme_bw() +
      ggplot2::theme(
        legend.key.width = grid::unit(1, "cm"),
        legend.position = legend_position,
        legend.direction = ifelse(
          legend_position %in% c("top", "bottom"),
          "horizontal",
          "vertical"
        )
      )
  }

  ############################################################# |
  # ---- Optionally, add table to the bottom of the plot. ----
  ############################################################# |
  if (!is.null(table)) {
    df_stats_table <- df_grp %>%
      dplyr::summarise(
        h_format_row(
          x = sfun(.data[[y]], ...)[table],
          format = table_format,
          labels = table_labels
        ),
        .groups = "drop"
      )

    stats_lev <- rev(setdiff(colnames(df_stats_table), c(group_var, x)))

    df_stats_table <- df_stats_table %>%
      tidyr::pivot_longer(
        cols = -dplyr::all_of(c(group_var, x)),
        names_to = "stat",
        values_to = "value",
        names_ptypes = list(stat = factor(levels = stats_lev))
      )

    tbl <- ggplot2::ggplot(
      df_stats_table,
      ggplot2::aes(x = .data[[x]], y = .data[["stat"]], label = .data[["value"]])
    ) +
      ggplot2::geom_text(size = table_font_size) +
      ggplot2::theme_bw() +
      ggplot2::theme(
        panel.border = ggplot2::element_blank(),
        panel.grid.major = ggplot2::element_blank(),
        panel.grid.minor = ggplot2::element_blank(),
        axis.ticks = ggplot2::element_blank(),
        axis.title = ggplot2::element_blank(),
        axis.text.x = ggplot2::element_blank(),
        axis.text.y = ggplot2::element_text(margin = ggplot2::margin(t = 0, r = 0, b = 0, l = 5)),
        strip.text = ggplot2::element_text(hjust = 0),
        strip.text.x = ggplot2::element_text(margin = ggplot2::margin(1.5, 0, 1.5, 0, "pt")),
        strip.background = ggplot2::element_rect(fill = "grey95", color = NA),
        legend.position = "none"
      )

    if (!is.null(group_var)) {
      tbl <- tbl + ggplot2::facet_wrap(facets = group_var, ncol = 1)
    }

    # align plot and table
    cowplot::plot_grid(p, tbl, ncol = 1, align = "v", axis = "tblr")
  } else {
    p
  }
}

#' Helper function to format the optional `g_lineplot` table
#'
#' @description `r lifecycle::badge("stable")`
#'
#' @param x (named `list`)\cr list of numerical values to be formatted and optionally labeled.
#'   Elements of `x` must be `numeric` vectors.
#' @param format (named `character` or `NULL`)\cr format patterns for `x`. Names of the `format` must
#'   match the names of `x`. This parameter is passed directly to the `rtables::format_rcell`
#'   function through the `format` parameter.
#' @param labels (named `character` or `NULL`)\cr optional labels for `x`. Names of the `labels` must
#'   match the names of `x`. When a label is not specified for an element of `x`,
#'   then this function tries to use `label` or `names` (in this order) attribute of that element
#'   (depending on which one exists and it is not `NULL` or `NA` or `NaN`). If none of these attributes
#'   are attached to a given element of `x`, then the label is automatically generated.
#'
#' @return A single row `data.frame` object.
#'
#' @examples
#' mean_ci <- c(48, 51)
#' x <- list(mean = 50, mean_ci = mean_ci)
#' format <- c(mean = "xx.x", mean_ci = "(xx.xx, xx.xx)")
#' labels <- c(mean = "My Mean")
#' h_format_row(x, format, labels)
#'
#' attr(mean_ci, "label") <- "Mean 95% CI"
#' x <- list(mean = 50, mean_ci = mean_ci)
#' h_format_row(x, format, labels)
#'
#' @export
h_format_row <- function(x, format, labels = NULL) {
  # cell: one row, one column data.frame
  format_cell <- function(x, format, label = NULL) {
    fc <- format_rcell(x = x, format = unlist(format))
    if (is.na(fc)) {
      fc <- "NA"
    }
    x_label <- attr(x, "label")
    if (!is.null(label) && !is.na(label)) {
      names(fc) <- label
    } else if (!is.null(x_label) && !is.na(x_label)) {
      names(fc) <- x_label
    } else if (length(x) == length(fc)) {
      names(fc) <- names(x)
    }
    as.data.frame(t(fc))
  }

  row <- do.call(
    cbind,
    lapply(
      names(x), function(xn) format_cell(x[[xn]], format = format[xn], label = labels[xn])
    )
  )

  row
}

#' Control function for `g_lineplot()`
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Default values for `variables` parameter in `g_lineplot` function.
#' A variable's default value can be overwritten for any variable.
#'
#' @param x (`string`)\cr x-variable name.
#' @param y (`string`)\cr y-variable name.
#' @param group_var (`string` or `NA`)\cr group variable name.
#' @param subject_var (`string` or `NA`)\cr subject variable name.
#' @param facet_var (`string` or `NA`)\cr faceting variable name.
#' @param paramcd (`string` or `NA`)\cr parameter code variable name.
#' @param y_unit (`string` or `NA`)\cr y-axis unit variable name.
#'
#' @return A named character vector of variable names.
#'
#' @examples
#' control_lineplot_vars()
#' control_lineplot_vars(group_var = NA)
#'
#' @export
control_lineplot_vars <- function(x = "AVISIT",
                                  y = "AVAL",
                                  group_var = "ARM",
                                  facet_var = NA,
                                  paramcd = "PARAMCD",
                                  y_unit = "AVALU",
                                  subject_var = "USUBJID") {
  checkmate::assert_string(x)
  checkmate::assert_string(y)
  checkmate::assert_string(group_var, na.ok = TRUE, null.ok = TRUE)
  checkmate::assert_string(facet_var, na.ok = TRUE, null.ok = TRUE)
  checkmate::assert_string(subject_var, na.ok = TRUE, null.ok = TRUE)
  checkmate::assert_string(paramcd, na.ok = TRUE, null.ok = TRUE)
  checkmate::assert_string(y_unit, na.ok = TRUE, null.ok = TRUE)

  variables <- c(
    x = x, y = y, group_var = group_var, paramcd = paramcd,
    y_unit = y_unit, subject_var = subject_var, facet_var = facet_var
  )
  return(variables)
}

#' Count the number of patients with particular flags
#'
#' @description `r lifecycle::badge("stable")`
#'
#' The analyze function [count_patients_with_flags()] creates a layout element to calculate counts of patients for
#' which user-specified flags are present.
#'
#' This function analyzes primary analysis variable `var` which indicates unique subject identifiers. Flags
#' variables to analyze are specified by the user via the `flag_variables` argument, and must either take value
#' `TRUE` (flag present) or `FALSE` (flag absent) for each record.
#'
#' If there are multiple records with the same flag present for a patient, only one occurrence is counted.
#'
#' @inheritParams argument_convention
#' @param flag_variables (`character`)\cr a vector specifying the names of `logical` variables from analysis dataset
#'   used for counting the number of unique identifiers.
#' @param flag_labels (`character`)\cr vector of labels to use for flag variables.
#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("count_patients_with_flags")`
#'   to see available statistics for this function.
#'
#' @seealso [count_patients_with_event]
#'
#' @name count_patients_with_flags
#' @order 1
NULL

#' @describeIn count_patients_with_flags Statistics function which counts the number of patients for which
#'   a particular flag variable is `TRUE`.
#'
#' @inheritParams analyze_variables
#' @param .var (`string`)\cr name of the column that contains the unique identifier.
#'
#' @note If `flag_labels` is not specified, variables labels will be extracted from `df`. If variables are not
#'   labeled, variable names will be used instead. Alternatively, a named `vector` can be supplied to
#'   `flag_variables` such that within each name-value pair the name corresponds to the variable name and the value is
#'   the label to use for this variable.
#'
#' @return
#' * `s_count_patients_with_flags()` returns the count and the fraction of unique identifiers with each particular
#'   flag as a list of statistics `n`, `count`, `count_fraction`, and `n_blq`, with one element per flag.
#'
#' @examples
#' # `s_count_patients_with_flags()`
#'
#' s_count_patients_with_flags(
#'   adae,
#'   "SUBJID",
#'   flag_variables = c("fl1", "fl2", "fl3", "fl4"),
#'   denom = "N_col",
#'   .N_col = 1000
#' )
#'
#' @export
s_count_patients_with_flags <- function(df,
                                        .var,
                                        flag_variables,
                                        flag_labels = NULL,
                                        .N_col, # nolint
                                        .N_row, # nolint
                                        denom = c("n", "N_row", "N_col")) {
  checkmate::assert_character(flag_variables)
  if (!is.null(flag_labels)) {
    checkmate::assert_character(flag_labels, len = length(flag_variables), any.missing = FALSE)
    flag_names <- flag_labels
  } else {
    if (is.null(names(flag_variables))) {
      flag_names <- formatters::var_labels(df[flag_variables], fill = TRUE)
    } else {
      flag_names <- unname(flag_variables)
      flag_variables <- names(flag_variables)
    }
  }

  checkmate::assert_subset(flag_variables, colnames(df))
  temp <- sapply(flag_variables, function(x) {
    tmp <- Map(function(y) which(df[[y]]), x)
    position_satisfy_flags <- Reduce(intersect, tmp)
    id_satisfy_flags <- as.character(unique(df[position_satisfy_flags, ][[.var]]))
    s_count_values(
      as.character(unique(df[[.var]])),
      id_satisfy_flags,
      denom = denom,
      .N_col = .N_col,
      .N_row = .N_row
    )
  })
  colnames(temp) <- flag_names
  temp <- data.frame(t(temp))
  result <- temp %>% as.list()
  if (length(flag_variables) == 1) {
    for (i in 1:3) names(result[[i]]) <- flag_names[1]
  }
  result
}

#' @describeIn count_patients_with_flags Formatted analysis function which is used as `afun`
#'   in `count_patients_with_flags()`.
#'
#' @return
#' * `a_count_patients_with_flags()` returns the corresponding list with formatted [rtables::CellValue()].
#'
#' @examples
#' #  We need to ungroup `count_fraction` first so that the `rtables` formatting
#' # function `format_count_fraction()` can be applied correctly.
#'
#' # `a_count_patients_with_flags()`
#'
#' afun <- make_afun(a_count_patients_with_flags,
#'   .stats = "count_fraction",
#'   .ungroup_stats = "count_fraction"
#' )
#' afun(
#'   adae,
#'   .N_col = 10L,
#'   .N_row = 10L,
#'   .var = "USUBJID",
#'   flag_variables = c("fl1", "fl2", "fl3", "fl4")
#' )
#'
#' @export
a_count_patients_with_flags <- make_afun(
  s_count_patients_with_flags,
  .formats = c("count_fraction" = format_count_fraction_fixed_dp)
)

#' @describeIn count_patients_with_flags Layout-creating function which can take statistics function
#'   arguments and additional format arguments. This function is a wrapper for [rtables::analyze()].
#'
#' @return
#' * `count_patients_with_flags()` returns a layout object suitable for passing to further layouting functions,
#'   or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
#'   the statistics from `s_count_patients_with_flags()` to the table layout.
#'
#' @examples
#' library(dplyr)
#'
#' # Add labelled flag variables to analysis dataset.
#' adae <- tern_ex_adae %>%
#'   mutate(
#'     fl1 = TRUE %>% with_label("Total AEs"),
#'     fl2 = (TRTEMFL == "Y") %>%
#'       with_label("Total number of patients with at least one adverse event"),
#'     fl3 = (TRTEMFL == "Y" & AEOUT == "FATAL") %>%
#'       with_label("Total number of patients with fatal AEs"),
#'     fl4 = (TRTEMFL == "Y" & AEOUT == "FATAL" & AEREL == "Y") %>%
#'       with_label("Total number of patients with related fatal AEs")
#'   )
#'
#' # `count_patients_with_flags()`
#'
#' lyt2 <- basic_table() %>%
#'   split_cols_by("ARM") %>%
#'   add_colcounts() %>%
#'   count_patients_with_flags(
#'     "SUBJID",
#'     flag_variables = c("fl1", "fl2", "fl3", "fl4"),
#'     denom = "N_col"
#'   )
#'
#' build_table(lyt2, adae, alt_counts_df = tern_ex_adsl)
#'
#' @export
#' @order 2
count_patients_with_flags <- function(lyt,
                                      var,
                                      flag_variables,
                                      flag_labels = NULL,
                                      var_labels = var,
                                      show_labels = "hidden",
                                      riskdiff = FALSE,
                                      na_str = default_na_str(),
                                      nested = TRUE,
                                      ...,
                                      table_names = paste0("tbl_flags_", var),
                                      .stats = "count_fraction",
                                      .formats = NULL,
                                      .indent_mods = NULL) {
  checkmate::assert_flag(riskdiff)

  s_args <- list(flag_variables = flag_variables, flag_labels = flag_labels, ...)

  afun <- make_afun(
    a_count_patients_with_flags,
    .stats = .stats,
    .formats = .formats,
    .indent_mods = .indent_mods,
    .ungroup_stats = .stats
  )

  extra_args <- if (isFALSE(riskdiff)) {
    s_args
  } else {
    list(
      afun = list("s_count_patients_with_flags" = afun),
      .stats = .stats,
      .indent_mods = .indent_mods,
      s_args = s_args
    )
  }

  lyt <- analyze(
    lyt = lyt,
    vars = var,
    var_labels = var_labels,
    show_labels = show_labels,
    afun = ifelse(isFALSE(riskdiff), afun, afun_riskdiff),
    table_names = table_names,
    na_str = na_str,
    nested = nested,
    extra_args = extra_args
  )

  lyt
}

#' Count occurrences
#'
#' @description `r lifecycle::badge("stable")`
#'
#' The analyze function [count_occurrences()] creates a layout element to calculate occurrence counts for patients.
#'
#' This function analyzes the variable(s) supplied to `vars` and returns a table of occurrence counts for
#' each unique value (or level) of the variable(s). This variable (or variables) must be
#' non-numeric. The `id` variable is used to indicate unique subject identifiers (defaults to `USUBJID`).
#'
#' If there are multiple occurrences of the same value recorded for a patient, the value is only counted once.
#'
#' The summarize function [summarize_occurrences()] performs the same function as [count_occurrences()] except it
#' creates content rows, not data rows, to summarize the current table row/column context and operates on the level of
#' the latest row split or the root of the table if no row splits have occurred.
#'
#' @inheritParams argument_convention
#' @param drop (`flag`)\cr whether non-appearing occurrence levels should be dropped from the resulting table.
#'   Note that in that case the remaining occurrence levels in the table are sorted alphabetically.
#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("count_occurrences")`
#'   to see available statistics for this function.
#'
#' @note By default, occurrences which don't appear in a given row split are dropped from the table and
#'   the occurrences in the table are sorted alphabetically per row split. Therefore, the corresponding layout
#'   needs to use `split_fun = drop_split_levels` in the `split_rows_by` calls. Use `drop = FALSE` if you would
#'   like to show all occurrences.
#'
#' @examples
#' library(dplyr)
#' df <- data.frame(
#'   USUBJID = as.character(c(
#'     1, 1, 2, 4, 4, 4,
#'     6, 6, 6, 7, 7, 8
#'   )),
#'   MHDECOD = c(
#'     "MH1", "MH2", "MH1", "MH1", "MH1", "MH3",
#'     "MH2", "MH2", "MH3", "MH1", "MH2", "MH4"
#'   ),
#'   ARM = rep(c("A", "B"), each = 6),
#'   SEX = c("F", "F", "M", "M", "M", "M", "F", "F", "F", "M", "M", "F")
#' )
#' df_adsl <- df %>%
#'   select(USUBJID, ARM) %>%
#'   unique()
#'
#' @name count_occurrences
#' @order 1
NULL

#' @describeIn count_occurrences Statistics function which counts number of patients that report an
#' occurrence.
#'
#' @param denom (`string`)\cr choice of denominator for patient proportions. Can be:
#'   - `N_col`: total number of patients in this column across rows
#'   - `n`: number of patients with any occurrences
#'
#' @return
#' * `s_count_occurrences()` returns a list with:
#'   * `count`: list of counts with one element per occurrence.
#'   * `count_fraction`: list of counts and fractions with one element per occurrence.
#'   * `fraction`: list of numerators and denominators with one element per occurrence.
#'
#' @examples
#' # Count unique occurrences per subject.
#' s_count_occurrences(
#'   df,
#'   .N_col = 4L,
#'   .df_row = df,
#'   .var = "MHDECOD",
#'   id = "USUBJID"
#' )
#'
#' @export
s_count_occurrences <- function(df,
                                denom = c("N_col", "n"),
                                .N_col, # nolint
                                .df_row,
                                drop = TRUE,
                                .var = "MHDECOD",
                                id = "USUBJID") {
  checkmate::assert_flag(drop)
  assert_df_with_variables(df, list(range = .var, id = id))
  checkmate::assert_count(.N_col)
  checkmate::assert_multi_class(df[[.var]], classes = c("factor", "character"))
  checkmate::assert_multi_class(df[[id]], classes = c("factor", "character"))
  denom <- match.arg(denom)

  occurrences <- if (drop) {
    # Note that we don't try to preserve original level order here since a) that would required
    # more time to look up in large original levels and b) that would fail for character input variable.
    occurrence_levels <- sort(unique(.df_row[[.var]]))
    if (length(occurrence_levels) == 0) {
      stop(
        "no empty `.df_row` input allowed when `drop = TRUE`,",
        " please use `split_fun = drop_split_levels` in the `rtables` `split_rows_by` calls"
      )
    }
    factor(df[[.var]], levels = occurrence_levels)
  } else {
    df[[.var]]
  }
  ids <- factor(df[[id]])
  dn <- switch(denom,
    n = nlevels(ids),
    N_col = .N_col
  )
  has_occurrence_per_id <- table(occurrences, ids) > 0
  n_ids_per_occurrence <- as.list(rowSums(has_occurrence_per_id))
  list(
    count = n_ids_per_occurrence,
    count_fraction = lapply(
      n_ids_per_occurrence,
      function(i, denom) {
        if (i == 0 && denom == 0) {
          c(0, 0)
        } else {
          c(i, i / denom)
        }
      },
      denom = dn
    ),
    fraction = lapply(
      n_ids_per_occurrence,
      function(i, denom) c("num" = i, "denom" = denom),
      denom = dn
    )
  )
}

#' @describeIn count_occurrences Formatted analysis function which is used as `afun`
#'   in `count_occurrences()`.
#'
#' @return
#' * `a_count_occurrences()` returns the corresponding list with formatted [rtables::CellValue()].
#'
#' @examples
#' a_count_occurrences(
#'   df,
#'   .N_col = 4L,
#'   .df_row = df,
#'   .var = "MHDECOD",
#'   id = "USUBJID"
#' )
#'
#' @export
a_count_occurrences <- function(df,
                                labelstr = "",
                                id = "USUBJID",
                                denom = c("N_col", "n"),
                                drop = TRUE,
                                .N_col, # nolint
                                .var = NULL,
                                .df_row = NULL,
                                .stats = NULL,
                                .formats = NULL,
                                .labels = NULL,
                                .indent_mods = NULL,
                                na_str = default_na_str()) {
  denom <- match.arg(denom)
  x_stats <- s_count_occurrences(
    df = df, denom = denom, .N_col = .N_col, .df_row = .df_row, drop = drop, .var = .var, id = id
  )
  if (is.null(unlist(x_stats))) {
    return(NULL)
  }
  x_lvls <- names(x_stats[[1]])

  # Fill in with formatting defaults if needed
  .stats <- get_stats("count_occurrences", stats_in = .stats)
  .formats <- get_formats_from_stats(.stats, .formats)
  .labels <- get_labels_from_stats(.stats, .labels, row_nms = x_lvls)
  .indent_mods <- get_indents_from_stats(.stats, .indent_mods, row_nms = x_lvls)

  if ("count_fraction_fixed_dp" %in% .stats) x_stats[["count_fraction_fixed_dp"]] <- x_stats[["count_fraction"]]
  x_stats <- x_stats[.stats]

  # Ungroup statistics with values for each level of x
  x_ungrp <- ungroup_stats(x_stats, .formats, list(), list())
  x_stats <- x_ungrp[["x"]]
  .formats <- x_ungrp[[".formats"]]

  # Auto format handling
  .formats <- apply_auto_formatting(.formats, x_stats, .df_row, .var)

  in_rows(
    .list = x_stats,
    .formats = .formats,
    .names = .labels,
    .labels = .labels,
    .indent_mods = .indent_mods,
    .format_na_strs = na_str
  )
}

#' @describeIn count_occurrences Layout-creating function which can take statistics function arguments
#'   and additional format arguments. This function is a wrapper for [rtables::analyze()].
#'
#' @return
#' * `count_occurrences()` returns a layout object suitable for passing to further layouting functions,
#'   or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
#'   the statistics from `s_count_occurrences()` to the table layout.
#'
#' @examples
#' # Create table layout
#' lyt <- basic_table() %>%
#'   split_cols_by("ARM") %>%
#'   add_colcounts() %>%
#'   count_occurrences(vars = "MHDECOD", .stats = c("count_fraction"))
#'
#' # Apply table layout to data and produce `rtable` object
#' tbl <- lyt %>%
#'   build_table(df, alt_counts_df = df_adsl) %>%
#'   prune_table()
#'
#' tbl
#'
#' @export
#' @order 2
count_occurrences <- function(lyt,
                              vars,
                              id = "USUBJID",
                              drop = TRUE,
                              var_labels = vars,
                              show_labels = "hidden",
                              riskdiff = FALSE,
                              na_str = default_na_str(),
                              nested = TRUE,
                              ...,
                              table_names = vars,
                              .stats = "count_fraction_fixed_dp",
                              .formats = NULL,
                              .labels = NULL,
                              .indent_mods = NULL) {
  checkmate::assert_flag(riskdiff)

  extra_args <- list(
    .stats = .stats, .formats = .formats, .labels = .labels, .indent_mods = .indent_mods, na_str = na_str
  )
  s_args <- list(id = id, drop = drop, ...)

  if (isFALSE(riskdiff)) {
    extra_args <- c(extra_args, s_args)
  } else {
    extra_args <- c(
      extra_args,
      list(
        afun = list("s_count_occurrences" = a_count_occurrences),
        s_args = s_args
      )
    )
  }

  analyze(
    lyt = lyt,
    vars = vars,
    afun = ifelse(isFALSE(riskdiff), a_count_occurrences, afun_riskdiff),
    var_labels = var_labels,
    show_labels = show_labels,
    table_names = table_names,
    na_str = na_str,
    nested = nested,
    extra_args = extra_args
  )
}

#' @describeIn count_occurrences Layout-creating function which can take content function arguments
#'   and additional format arguments. This function is a wrapper for [rtables::summarize_row_groups()].
#'
#' @return
#' * `summarize_occurrences()` returns a layout object suitable for passing to further layouting functions,
#'   or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted content rows
#'   containing the statistics from `s_count_occurrences()` to the table layout.
#'
#' @examples
#' # Layout creating function with custom format.
#' basic_table() %>%
#'   add_colcounts() %>%
#'   split_rows_by("SEX", child_labels = "visible") %>%
#'   summarize_occurrences(
#'     var = "MHDECOD",
#'     .formats = c("count_fraction" = "xx.xx (xx.xx%)")
#'   ) %>%
#'   build_table(df, alt_counts_df = df_adsl)
#'
#' @export
#' @order 3
summarize_occurrences <- function(lyt,
                                  var,
                                  id = "USUBJID",
                                  drop = TRUE,
                                  riskdiff = FALSE,
                                  na_str = default_na_str(),
                                  ...,
                                  .stats = "count_fraction_fixed_dp",
                                  .formats = NULL,
                                  .indent_mods = NULL,
                                  .labels = NULL) {
  checkmate::assert_flag(riskdiff)

  extra_args <- list(
    .stats = .stats, .formats = .formats, .labels = .labels, .indent_mods = .indent_mods, na_str = na_str
  )
  s_args <- list(id = id, drop = drop, ...)

  if (isFALSE(riskdiff)) {
    extra_args <- c(extra_args, s_args)
  } else {
    extra_args <- c(
      extra_args,
      list(
        afun = list("s_count_occurrences" = a_count_occurrences),
        s_args = s_args
      )
    )
  }

  summarize_row_groups(
    lyt = lyt,
    var = var,
    cfun = ifelse(isFALSE(riskdiff), a_count_occurrences, afun_riskdiff),
    na_str = na_str,
    extra_args = extra_args
  )
}

#' Tabulate biomarker effects on binary response by subgroup
#'
#' @description `r lifecycle::badge("stable")`
#'
#' The [tabulate_rsp_biomarkers()] function creates a layout element to tabulate the estimated biomarker effects on a
#' binary response endpoint across subgroups, returning statistics including response rate and odds ratio for each
#' population subgroup. The table is created from `df`, a list of data frames returned by [extract_rsp_biomarkers()],
#' with the statistics to include specified via the `vars` parameter.
#'
#' A forest plot can be created from the resulting table using the [g_forest()] function.
#'
#' @inheritParams argument_convention
#' @param df (`data.frame`)\cr containing all analysis variables, as returned by
#'   [extract_rsp_biomarkers()].
#' @param vars (`character`)\cr the names of statistics to be reported among:
#'   * `n_tot`: Total number of patients per group.
#'   * `n_rsp`: Total number of responses per group.
#'   * `prop`: Total response proportion per group.
#'   * `or`: Odds ratio.
#'   * `ci`: Confidence interval of odds ratio.
#'   * `pval`: p-value of the effect.
#'   Note, the statistics `n_tot`, `or` and `ci` are required.
#'
#' @return An `rtables` table summarizing biomarker effects on binary response by subgroup.
#'
#' @details These functions create a layout starting from a data frame which contains
#'   the required statistics. The tables are then typically used as input for forest plots.
#'
#' @note In contrast to [tabulate_rsp_subgroups()] this tabulation function does
#'   not start from an input layout `lyt`. This is because internally the table is
#'   created by combining multiple subtables.
#'
#' @seealso [h_tab_rsp_one_biomarker()] which is used internally, [extract_rsp_biomarkers()].
#'
#' @examples
#' library(dplyr)
#' library(forcats)
#'
#' adrs <- tern_ex_adrs
#' adrs_labels <- formatters::var_labels(adrs)
#'
#' adrs_f <- adrs %>%
#'   filter(PARAMCD == "BESRSPI") %>%
#'   mutate(rsp = AVALC == "CR")
#' formatters::var_labels(adrs_f) <- c(adrs_labels, "Response")
#'
#' df <- extract_rsp_biomarkers(
#'   variables = list(
#'     rsp = "rsp",
#'     biomarkers = c("BMRKR1", "AGE"),
#'     covariates = "SEX",
#'     subgroups = "BMRKR2"
#'   ),
#'   data = adrs_f
#' )
#'
#' \donttest{
#' ## Table with default columns.
#' tabulate_rsp_biomarkers(df)
#'
#' ## Table with a manually chosen set of columns: leave out "pval", reorder.
#' tab <- tabulate_rsp_biomarkers(
#'   df = df,
#'   vars = c("n_rsp", "ci", "n_tot", "prop", "or")
#' )
#'
#' ## Finally produce the forest plot.
#' g_forest(tab, xlim = c(0.7, 1.4))
#' }
#'
#' @export
#' @name response_biomarkers_subgroups
tabulate_rsp_biomarkers <- function(df,
                                    vars = c("n_tot", "n_rsp", "prop", "or", "ci", "pval"),
                                    na_str = default_na_str(),
                                    .indent_mods = 0L) {
  checkmate::assert_data_frame(df)
  checkmate::assert_character(df$biomarker)
  checkmate::assert_character(df$biomarker_label)
  checkmate::assert_subset(vars, get_stats("tabulate_rsp_biomarkers"))

  # Create "ci" column from "lcl" and "ucl"
  df$ci <- combine_vectors(df$lcl, df$ucl)

  df_subs <- split(df, f = df$biomarker)
  tabs <- lapply(df_subs, FUN = function(df_sub) {
    tab_sub <- h_tab_rsp_one_biomarker(
      df = df_sub,
      vars = vars,
      na_str = na_str,
      .indent_mods = .indent_mods
    )
    # Insert label row as first row in table.
    label_at_path(tab_sub, path = row_paths(tab_sub)[[1]][1]) <- df_sub$biomarker_label[1]
    tab_sub
  })
  result <- do.call(rbind, tabs)

  n_id <- grep("n_tot", vars)
  or_id <- match("or", vars)
  ci_id <- match("ci", vars)
  structure(
    result,
    forest_header = paste0(c("Lower", "Higher"), "\nBetter"),
    col_x = or_id,
    col_ci = ci_id,
    col_symbol_size = n_id
  )
}

#' Prepare response data estimates for multiple biomarkers in a single data frame
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Prepares estimates for number of responses, patients and overall response rate,
#' as well as odds ratio estimates, confidence intervals and p-values,
#' for multiple biomarkers across population subgroups in a single data frame.
#' `variables` corresponds to the names of variables found in `data`, passed as a
#' named list and requires elements `rsp` and `biomarkers` (vector of continuous
#' biomarker variables) and optionally `covariates`, `subgroups` and `strata`.
#' `groups_lists` optionally specifies groupings for `subgroups` variables.
#'
#' @inheritParams argument_convention
#' @inheritParams response_subgroups
#' @param control (named `list`)\cr controls for the response definition and the
#'   confidence level produced by [control_logistic()].
#'
#' @return A `data.frame` with columns `biomarker`, `biomarker_label`, `n_tot`, `n_rsp`,
#'   `prop`, `or`, `lcl`, `ucl`, `conf_level`, `pval`, `pval_label`, `subgroup`, `var`,
#'   `var_label`, and `row_type`.
#'
#' @note You can also specify a continuous variable in `rsp` and then use the
#'   `response_definition` control to convert that internally to a logical
#'   variable reflecting binary response.
#'
#' @seealso [h_logistic_mult_cont_df()] which is used internally.
#'
#' @examples
#' library(dplyr)
#' library(forcats)
#'
#' adrs <- tern_ex_adrs
#' adrs_labels <- formatters::var_labels(adrs)
#'
#' adrs_f <- adrs %>%
#'   filter(PARAMCD == "BESRSPI") %>%
#'   mutate(rsp = AVALC == "CR")
#'
#' # Typical analysis of two continuous biomarkers `BMRKR1` and `AGE`,
#' # in logistic regression models with one covariate `RACE`. The subgroups
#' # are defined by the levels of `BMRKR2`.
#' df <- extract_rsp_biomarkers(
#'   variables = list(
#'     rsp = "rsp",
#'     biomarkers = c("BMRKR1", "AGE"),
#'     covariates = "SEX",
#'     subgroups = "BMRKR2"
#'   ),
#'   data = adrs_f
#' )
#' df
#'
#' # Here we group the levels of `BMRKR2` manually, and we add a stratification
#' # variable `STRATA1`. We also here use a continuous variable `EOSDY`
#' # which is then binarized internally (response is defined as this variable
#' # being larger than 750).
#' df_grouped <- extract_rsp_biomarkers(
#'   variables = list(
#'     rsp = "EOSDY",
#'     biomarkers = c("BMRKR1", "AGE"),
#'     covariates = "SEX",
#'     subgroups = "BMRKR2",
#'     strata = "STRATA1"
#'   ),
#'   data = adrs_f,
#'   groups_lists = list(
#'     BMRKR2 = list(
#'       "low" = "LOW",
#'       "low/medium" = c("LOW", "MEDIUM"),
#'       "low/medium/high" = c("LOW", "MEDIUM", "HIGH")
#'     )
#'   ),
#'   control = control_logistic(
#'     response_definition = "I(response > 750)"
#'   )
#' )
#' df_grouped
#'
#' @export
extract_rsp_biomarkers <- function(variables,
                                   data,
                                   groups_lists = list(),
                                   control = control_logistic(),
                                   label_all = "All Patients") {
  if ("strat" %in% names(variables)) {
    warning(
      "Warning: the `strat` element name of the `variables` list argument to `extract_rsp_biomarkers() ",
      "was deprecated in tern 0.9.4.\n  ",
      "Please use the name `strata` instead of `strat` in the `variables` argument."
    )
    variables[["strata"]] <- variables[["strat"]]
  }

  assert_list_of_variables(variables)
  checkmate::assert_string(variables$rsp)
  checkmate::assert_character(variables$subgroups, null.ok = TRUE)
  checkmate::assert_string(label_all)

  # Start with all patients.
  result_all <- h_logistic_mult_cont_df(
    variables = variables,
    data = data,
    control = control
  )
  result_all$subgroup <- label_all
  result_all$var <- "ALL"
  result_all$var_label <- label_all
  result_all$row_type <- "content"
  if (is.null(variables$subgroups)) {
    # Only return result for all patients.
    result_all
  } else {
    # Add subgroups results.
    l_data <- h_split_by_subgroups(
      data,
      variables$subgroups,
      groups_lists = groups_lists
    )
    l_result <- lapply(l_data, function(grp) {
      result <- h_logistic_mult_cont_df(
        variables = variables,
        data = grp$df,
        control = control
      )
      result_labels <- grp$df_labels[rep(1, times = nrow(result)), ]
      cbind(result, result_labels)
    })
    result_subgroups <- do.call(rbind, args = c(l_result, make.row.names = FALSE))
    result_subgroups$row_type <- "analysis"
    rbind(
      result_all,
      result_subgroups
    )
  }
}

#' Multivariate logistic regression table
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Layout-creating function which summarizes a logistic variable regression for binary outcome with
#' categorical/continuous covariates in model statement. For each covariate category (if categorical)
#' or specified values (if continuous), present degrees of freedom, regression parameter estimate and
#' standard error (SE) relative to reference group or category. Report odds ratios for each covariate
#' category or specified values and corresponding Wald confidence intervals as default but allow user
#' to specify other confidence levels. Report p-value for Wald chi-square test of the null hypothesis
#' that covariate has no effect on response in model containing all specified covariates.
#' Allow option to include one two-way interaction and present similar output for
#' each interaction degree of freedom.
#'
#' @inheritParams argument_convention
#' @param drop_and_remove_str (`string`)\cr string to be dropped and removed.
#'
#' @return A layout object suitable for passing to further layouting functions, or to [rtables::build_table()].
#'   Adding this function to an `rtable` layout will add a logistic regression variable summary to the table layout.
#'
#' @note For the formula, the variable names need to be standard `data.frame` column names without
#'   special characters.
#'
#' @examples
#' library(dplyr)
#' library(broom)
#'
#' adrs_f <- tern_ex_adrs %>%
#'   filter(PARAMCD == "BESRSPI") %>%
#'   filter(RACE %in% c("ASIAN", "WHITE", "BLACK OR AFRICAN AMERICAN")) %>%
#'   mutate(
#'     Response = case_when(AVALC %in% c("PR", "CR") ~ 1, TRUE ~ 0),
#'     RACE = factor(RACE),
#'     SEX = factor(SEX)
#'   )
#' formatters::var_labels(adrs_f) <- c(formatters::var_labels(tern_ex_adrs), Response = "Response")
#' mod1 <- fit_logistic(
#'   data = adrs_f,
#'   variables = list(
#'     response = "Response",
#'     arm = "ARMCD",
#'     covariates = c("AGE", "RACE")
#'   )
#' )
#' mod2 <- fit_logistic(
#'   data = adrs_f,
#'   variables = list(
#'     response = "Response",
#'     arm = "ARMCD",
#'     covariates = c("AGE", "RACE"),
#'     interaction = "AGE"
#'   )
#' )
#'
#' df <- tidy(mod1, conf_level = 0.99)
#' df2 <- tidy(mod2, conf_level = 0.99)
#'
#' # flagging empty strings with "_"
#' df <- df_explicit_na(df, na_level = "_")
#' df2 <- df_explicit_na(df2, na_level = "_")
#'
#' result1 <- basic_table() %>%
#'   summarize_logistic(
#'     conf_level = 0.95,
#'     drop_and_remove_str = "_"
#'   ) %>%
#'   build_table(df = df)
#' result1
#'
#' result2 <- basic_table() %>%
#'   summarize_logistic(
#'     conf_level = 0.95,
#'     drop_and_remove_str = "_"
#'   ) %>%
#'   build_table(df = df2)
#' result2
#'
#' @export
#' @order 1
summarize_logistic <- function(lyt,
                               conf_level,
                               drop_and_remove_str = "",
                               .indent_mods = NULL) {
  # checks
  checkmate::assert_string(drop_and_remove_str)

  sum_logistic_variable_test <- logistic_summary_by_flag("is_variable_summary")
  sum_logistic_term_estimates <- logistic_summary_by_flag("is_term_summary", .indent_mods = .indent_mods)
  sum_logistic_odds_ratios <- logistic_summary_by_flag("is_reference_summary", .indent_mods = .indent_mods)
  split_fun <- drop_and_remove_levels(drop_and_remove_str)

  lyt <- logistic_regression_cols(lyt, conf_level = conf_level)
  lyt <- split_rows_by(lyt, var = "variable", labels_var = "variable_label", split_fun = split_fun)
  lyt <- sum_logistic_variable_test(lyt)
  lyt <- split_rows_by(lyt, var = "term", labels_var = "term_label", split_fun = split_fun)
  lyt <- sum_logistic_term_estimates(lyt)
  lyt <- split_rows_by(lyt, var = "interaction", labels_var = "interaction_label", split_fun = split_fun)
  lyt <- split_rows_by(lyt, var = "reference", labels_var = "reference_label", split_fun = split_fun)
  lyt <- sum_logistic_odds_ratios(lyt)
  lyt
}

#' Fit for logistic regression
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Fit a (conditional) logistic regression model.
#'
#' @inheritParams argument_convention
#' @param data (`data.frame`)\cr the data frame on which the model was fit.
#' @param response_definition (`string`)\cr the definition of what an event is in terms of `response`.
#'   This will be used when fitting the (conditional) logistic regression model on the left hand
#'   side of the formula.
#'
#' @return A fitted logistic regression model.
#'
#' @section Model Specification:
#'
#' The `variables` list needs to include the following elements:
#'   * `arm`: Treatment arm variable name.
#'   * `response`: The response arm variable name. Usually this is a 0/1 variable.
#'   * `covariates`: This is either `NULL` (no covariates) or a character vector of covariate variable names.
#'   * `interaction`: This is either `NULL` (no interaction) or a string of a single covariate variable name already
#'     included in `covariates`. Then the interaction with the treatment arm is included in the model.
#'
#' @examples
#' library(dplyr)
#'
#' adrs_f <- tern_ex_adrs %>%
#'   filter(PARAMCD == "BESRSPI") %>%
#'   filter(RACE %in% c("ASIAN", "WHITE", "BLACK OR AFRICAN AMERICAN")) %>%
#'   mutate(
#'     Response = case_when(AVALC %in% c("PR", "CR") ~ 1, TRUE ~ 0),
#'     RACE = factor(RACE),
#'     SEX = factor(SEX)
#'   )
#' formatters::var_labels(adrs_f) <- c(formatters::var_labels(tern_ex_adrs), Response = "Response")
#' mod1 <- fit_logistic(
#'   data = adrs_f,
#'   variables = list(
#'     response = "Response",
#'     arm = "ARMCD",
#'     covariates = c("AGE", "RACE")
#'   )
#' )
#' mod2 <- fit_logistic(
#'   data = adrs_f,
#'   variables = list(
#'     response = "Response",
#'     arm = "ARMCD",
#'     covariates = c("AGE", "RACE"),
#'     interaction = "AGE"
#'   )
#' )
#'
#' @export
fit_logistic <- function(data,
                         variables = list(
                           response = "Response",
                           arm = "ARMCD",
                           covariates = NULL,
                           interaction = NULL,
                           strata = NULL
                         ),
                         response_definition = "response") {
  assert_df_with_variables(data, variables)
  checkmate::assert_subset(names(variables), c("response", "arm", "covariates", "interaction", "strata"))
  checkmate::assert_string(response_definition)
  checkmate::assert_true(grepl("response", response_definition))

  response_definition <- sub(
    pattern = "response",
    replacement = variables$response,
    x = response_definition,
    fixed = TRUE
  )
  form <- paste0(response_definition, " ~ ", variables$arm)
  if (!is.null(variables$covariates)) {
    form <- paste0(form, " + ", paste(variables$covariates, collapse = " + "))
  }
  if (!is.null(variables$interaction)) {
    checkmate::assert_string(variables$interaction)
    checkmate::assert_subset(variables$interaction, variables$covariates)
    form <- paste0(form, " + ", variables$arm, ":", variables$interaction)
  }
  if (!is.null(variables$strata)) {
    strata_arg <- if (length(variables$strata) > 1) {
      paste0("I(interaction(", paste0(variables$strata, collapse = ", "), "))")
    } else {
      variables$strata
    }
    form <- paste0(form, "+ strata(", strata_arg, ")")
  }
  formula <- stats::as.formula(form)
  if (is.null(variables$strata)) {
    stats::glm(
      formula = formula,
      data = data,
      family = stats::binomial("logit")
    )
  } else {
    clogit_with_tryCatch(
      formula = formula,
      data = data,
      x = TRUE
    )
  }
}

#' Custom tidy method for binomial GLM results
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Helper method (for [broom::tidy()]) to prepare a data frame from a `glm` object
#' with `binomial` family.
#'
#' @inheritParams argument_convention
#' @param at (`numeric` or `NULL`)\cr optional values for the interaction variable. Otherwise the median is used.
#' @param x (`glm`)\cr logistic regression model fitted by [stats::glm()] with "binomial" family.
#'
#' @return A `data.frame` containing the tidied model.
#'
#' @method tidy glm
#'
#' @seealso [h_logistic_regression] for relevant helper functions.
#'
#' @examples
#' library(dplyr)
#' library(broom)
#'
#' adrs_f <- tern_ex_adrs %>%
#'   filter(PARAMCD == "BESRSPI") %>%
#'   filter(RACE %in% c("ASIAN", "WHITE", "BLACK OR AFRICAN AMERICAN")) %>%
#'   mutate(
#'     Response = case_when(AVALC %in% c("PR", "CR") ~ 1, TRUE ~ 0),
#'     RACE = factor(RACE),
#'     SEX = factor(SEX)
#'   )
#' formatters::var_labels(adrs_f) <- c(formatters::var_labels(tern_ex_adrs), Response = "Response")
#' mod1 <- fit_logistic(
#'   data = adrs_f,
#'   variables = list(
#'     response = "Response",
#'     arm = "ARMCD",
#'     covariates = c("AGE", "RACE")
#'   )
#' )
#' mod2 <- fit_logistic(
#'   data = adrs_f,
#'   variables = list(
#'     response = "Response",
#'     arm = "ARMCD",
#'     covariates = c("AGE", "RACE"),
#'     interaction = "AGE"
#'   )
#' )
#'
#' df <- tidy(mod1, conf_level = 0.99)
#' df2 <- tidy(mod2, conf_level = 0.99)
#'
#' @export
tidy.glm <- function(x, # nolint
                     conf_level = 0.95,
                     at = NULL,
                     ...) {
  checkmate::assert_class(x, "glm")
  checkmate::assert_set_equal(x$family$family, "binomial")

  terms_name <- attr(stats::terms(x), "term.labels")
  xs_class <- attr(x$terms, "dataClasses")
  interaction <- terms_name[which(!terms_name %in% names(xs_class))]
  df <- if (length(interaction) == 0) {
    h_logistic_simple_terms(
      x = terms_name,
      fit_glm = x,
      conf_level = conf_level
    )
  } else {
    h_logistic_inter_terms(
      x = terms_name,
      fit_glm = x,
      conf_level = conf_level,
      at = at
    )
  }
  for (var in c("variable", "term", "interaction", "reference")) {
    df[[var]] <- factor(df[[var]], levels = unique(df[[var]]))
  }
  df
}

#' Logistic regression multivariate column layout function
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Layout-creating function which creates a multivariate column layout summarizing logistic
#' regression results. This function is a wrapper for [rtables::split_cols_by_multivar()].
#'
#' @inheritParams argument_convention
#'
#' @return A layout object suitable for passing to further layouting functions. Adding this
#'   function to an `rtable` layout will split the table into columns corresponding to
#'   statistics `df`, `estimate`, `std_error`, `odds_ratio`, `ci`, and `pvalue`.
#'
#' @export
logistic_regression_cols <- function(lyt,
                                     conf_level = 0.95) {
  vars <- c("df", "estimate", "std_error", "odds_ratio", "ci", "pvalue")
  var_labels <- c(
    df = "Degrees of Freedom",
    estimate = "Parameter Estimate",
    std_error = "Standard Error",
    odds_ratio = "Odds Ratio",
    ci = paste("Wald", f_conf_level(conf_level)),
    pvalue = "p-value"
  )
  split_cols_by_multivar(
    lyt = lyt,
    vars = vars,
    varlabels = var_labels
  )
}

#' Logistic regression summary table
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Constructor for content functions to be used in [`summarize_logistic()`] to summarize
#' logistic regression results. This function is a wrapper for [rtables::summarize_row_groups()].
#'
#' @inheritParams argument_convention
#' @param flag_var (`string`)\cr variable name identifying which row should be used in this
#'   content function.
#'
#' @return A content function.
#'
#' @export
logistic_summary_by_flag <- function(flag_var, na_str = default_na_str(), .indent_mods = NULL) {
  checkmate::assert_string(flag_var)
  function(lyt) {
    cfun_list <- list(
      df = cfun_by_flag("df", flag_var, format = "xx.", .indent_mods = .indent_mods),
      estimate = cfun_by_flag("estimate", flag_var, format = "xx.xxx", .indent_mods = .indent_mods),
      std_error = cfun_by_flag("std_error", flag_var, format = "xx.xxx", .indent_mods = .indent_mods),
      odds_ratio = cfun_by_flag("odds_ratio", flag_var, format = ">999.99", .indent_mods = .indent_mods),
      ci = cfun_by_flag("ci", flag_var, format = format_extreme_values_ci(2L), .indent_mods = .indent_mods),
      pvalue = cfun_by_flag("pvalue", flag_var, format = "x.xxxx | (<0.0001)", .indent_mods = .indent_mods)
    )
    summarize_row_groups(
      lyt = lyt,
      cfun = cfun_list,
      na_str = na_str
    )
  }
}

#' Helper functions for tabulating binary response by subgroup
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Helper functions that tabulate in a data frame statistics such as response rate
#' and odds ratio for population subgroups.
#'
#' @inheritParams argument_convention
#' @inheritParams response_subgroups
#' @param arm (`factor`)\cr the treatment group variable.
#'
#' @details Main functionality is to prepare data for use in a layout-creating function.
#'
#' @examples
#' library(dplyr)
#' library(forcats)
#'
#' adrs <- tern_ex_adrs
#' adrs_labels <- formatters::var_labels(adrs)
#'
#' adrs_f <- adrs %>%
#'   filter(PARAMCD == "BESRSPI") %>%
#'   filter(ARM %in% c("A: Drug X", "B: Placebo")) %>%
#'   droplevels() %>%
#'   mutate(
#'     # Reorder levels of factor to make the placebo group the reference arm.
#'     ARM = fct_relevel(ARM, "B: Placebo"),
#'     rsp = AVALC == "CR"
#'   )
#' formatters::var_labels(adrs_f) <- c(adrs_labels, "Response")
#'
#' @name h_response_subgroups
NULL

#' @describeIn h_response_subgroups Helper to prepare a data frame of binary responses by arm.
#'
#' @return
#' * `h_proportion_df()` returns a `data.frame` with columns `arm`, `n`, `n_rsp`, and `prop`.
#'
#' @examples
#' h_proportion_df(
#'   c(TRUE, FALSE, FALSE),
#'   arm = factor(c("A", "A", "B"), levels = c("A", "B"))
#' )
#'
#' @export
h_proportion_df <- function(rsp, arm) {
  checkmate::assert_logical(rsp)
  assert_valid_factor(arm, len = length(rsp))
  non_missing_rsp <- !is.na(rsp)
  rsp <- rsp[non_missing_rsp]
  arm <- arm[non_missing_rsp]

  lst_rsp <- split(rsp, arm)
  lst_results <- Map(function(x, arm) {
    if (length(x) > 0) {
      s_prop <- s_proportion(df = x)
      data.frame(
        arm = arm,
        n = length(x),
        n_rsp = unname(s_prop$n_prop[1]),
        prop = unname(s_prop$n_prop[2]),
        stringsAsFactors = FALSE
      )
    } else {
      data.frame(
        arm = arm,
        n = 0L,
        n_rsp = NA,
        prop = NA,
        stringsAsFactors = FALSE
      )
    }
  }, lst_rsp, names(lst_rsp))

  df <- do.call(rbind, args = c(lst_results, make.row.names = FALSE))
  df$arm <- factor(df$arm, levels = levels(arm))
  df
}

#' @describeIn h_response_subgroups Summarizes proportion of binary responses by arm and across subgroups
#'    in a data frame. `variables` corresponds to the names of variables found in `data`, passed as a named list and
#'    requires elements `rsp`, `arm` and optionally `subgroups`. `groups_lists` optionally specifies
#'    groupings for `subgroups` variables.
#'
#' @return
#' * `h_proportion_subgroups_df()` returns a `data.frame` with columns `arm`, `n`, `n_rsp`, `prop`, `subgroup`,
#'   `var`, `var_label`, and `row_type`.
#'
#' @examples
#' h_proportion_subgroups_df(
#'   variables = list(rsp = "rsp", arm = "ARM", subgroups = c("SEX", "BMRKR2")),
#'   data = adrs_f
#' )
#'
#' # Define groupings for BMRKR2 levels.
#' h_proportion_subgroups_df(
#'   variables = list(rsp = "rsp", arm = "ARM", subgroups = c("SEX", "BMRKR2")),
#'   data = adrs_f,
#'   groups_lists = list(
#'     BMRKR2 = list(
#'       "low" = "LOW",
#'       "low/medium" = c("LOW", "MEDIUM"),
#'       "low/medium/high" = c("LOW", "MEDIUM", "HIGH")
#'     )
#'   )
#' )
#'
#' @export
h_proportion_subgroups_df <- function(variables,
                                      data,
                                      groups_lists = list(),
                                      label_all = "All Patients") {
  checkmate::assert_character(variables$rsp)
  checkmate::assert_character(variables$arm)
  checkmate::assert_character(variables$subgroups, null.ok = TRUE)
  assert_df_with_factors(data, list(val = variables$arm), min.levels = 2, max.levels = 2)
  assert_df_with_variables(data, variables)
  checkmate::assert_string(label_all)

  # Add All Patients.
  result_all <- h_proportion_df(data[[variables$rsp]], data[[variables$arm]])
  result_all$subgroup <- label_all
  result_all$var <- "ALL"
  result_all$var_label <- label_all
  result_all$row_type <- "content"

  # Add Subgroups.
  if (is.null(variables$subgroups)) {
    result_all
  } else {
    l_data <- h_split_by_subgroups(data, variables$subgroups, groups_lists = groups_lists)

    l_result <- lapply(l_data, function(grp) {
      result <- h_proportion_df(grp$df[[variables$rsp]], grp$df[[variables$arm]])
      result_labels <- grp$df_labels[rep(1, times = nrow(result)), ]
      cbind(result, result_labels)
    })
    result_subgroups <- do.call(rbind, args = c(l_result, make.row.names = FALSE))
    result_subgroups$row_type <- "analysis"

    rbind(
      result_all,
      result_subgroups
    )
  }
}

#' @describeIn h_response_subgroups Helper to prepare a data frame with estimates of
#'   the odds ratio between a treatment and a control arm.
#'
#' @inheritParams response_subgroups
#' @param strata_data (`factor`, `data.frame`, or `NULL`)\cr required if stratified analysis is performed.
#'
#' @return
#' * `h_odds_ratio_df()` returns a `data.frame` with columns `arm`, `n_tot`, `or`, `lcl`, `ucl`, `conf_level`, and
#'   optionally `pval` and `pval_label`.
#'
#' @examples
#' # Unstratatified analysis.
#' h_odds_ratio_df(
#'   c(TRUE, FALSE, FALSE, TRUE),
#'   arm = factor(c("A", "A", "B", "B"), levels = c("A", "B"))
#' )
#'
#' # Include p-value.
#' h_odds_ratio_df(adrs_f$rsp, adrs_f$ARM, method = "chisq")
#'
#' # Stratatified analysis.
#' h_odds_ratio_df(
#'   rsp = adrs_f$rsp,
#'   arm = adrs_f$ARM,
#'   strata_data = adrs_f[, c("STRATA1", "STRATA2")],
#'   method = "cmh"
#' )
#'
#' @export
h_odds_ratio_df <- function(rsp, arm, strata_data = NULL, conf_level = 0.95, method = NULL) {
  assert_valid_factor(arm, n.levels = 2, len = length(rsp))

  df_rsp <- data.frame(
    rsp = rsp,
    arm = arm
  )

  if (!is.null(strata_data)) {
    strata_var <- interaction(strata_data, drop = TRUE)
    strata_name <- "strata"

    assert_valid_factor(strata_var, len = nrow(df_rsp))

    df_rsp[[strata_name]] <- strata_var
  } else {
    strata_name <- NULL
  }

  l_df <- split(df_rsp, arm)

  if (nrow(l_df[[1]]) > 0 && nrow(l_df[[2]]) > 0) {
    # Odds ratio and CI.
    result_odds_ratio <- s_odds_ratio(
      df = l_df[[2]],
      .var = "rsp",
      .ref_group = l_df[[1]],
      .in_ref_col = FALSE,
      .df_row = df_rsp,
      variables = list(arm = "arm", strata = strata_name),
      conf_level = conf_level
    )

    df <- data.frame(
      # Dummy column needed downstream to create a nested header.
      arm = " ",
      n_tot = unname(result_odds_ratio$n_tot["n_tot"]),
      or = unname(result_odds_ratio$or_ci["est"]),
      lcl = unname(result_odds_ratio$or_ci["lcl"]),
      ucl = unname(result_odds_ratio$or_ci["ucl"]),
      conf_level = conf_level,
      stringsAsFactors = FALSE
    )

    if (!is.null(method)) {
      # Test for difference.
      result_test <- s_test_proportion_diff(
        df = l_df[[2]],
        .var = "rsp",
        .ref_group = l_df[[1]],
        .in_ref_col = FALSE,
        variables = list(strata = strata_name),
        method = method
      )

      df$pval <- as.numeric(result_test$pval)
      df$pval_label <- obj_label(result_test$pval)
    }

    # In those cases cannot go through the model so will obtain n_tot from data.
  } else if (
    (nrow(l_df[[1]]) == 0 && nrow(l_df[[2]]) > 0) ||
      (nrow(l_df[[1]]) > 0 && nrow(l_df[[2]]) == 0)
  ) {
    df <- data.frame(
      # Dummy column needed downstream to create a nested header.
      arm = " ",
      n_tot = sum(stats::complete.cases(df_rsp)),
      or = NA,
      lcl = NA,
      ucl = NA,
      conf_level = conf_level,
      stringsAsFactors = FALSE
    )
    if (!is.null(method)) {
      df$pval <- NA
      df$pval_label <- NA
    }
  } else {
    df <- data.frame(
      # Dummy column needed downstream to create a nested header.
      arm = " ",
      n_tot = 0L,
      or = NA,
      lcl = NA,
      ucl = NA,
      conf_level = conf_level,
      stringsAsFactors = FALSE
    )

    if (!is.null(method)) {
      df$pval <- NA
      df$pval_label <- NA
    }
  }

  df
}

#' @describeIn h_response_subgroups Summarizes estimates of the odds ratio between a treatment and a control
#'   arm across subgroups in a data frame. `variables` corresponds to the names of variables found in
#'   `data`, passed as a named list and requires elements `rsp`, `arm` and optionally `subgroups`
#'   and `strata`. `groups_lists` optionally specifies groupings for `subgroups` variables.
#'
#' @return
#' * `h_odds_ratio_subgroups_df()` returns a `data.frame` with columns `arm`, `n_tot`, `or`, `lcl`, `ucl`,
#'   `conf_level`, `subgroup`, `var`, `var_label`, and `row_type`.
#'
#' @examples
#' # Unstratified analysis.
#' h_odds_ratio_subgroups_df(
#'   variables = list(rsp = "rsp", arm = "ARM", subgroups = c("SEX", "BMRKR2")),
#'   data = adrs_f
#' )
#'
#' # Stratified analysis.
#' h_odds_ratio_subgroups_df(
#'   variables = list(
#'     rsp = "rsp",
#'     arm = "ARM",
#'     subgroups = c("SEX", "BMRKR2"),
#'     strata = c("STRATA1", "STRATA2")
#'   ),
#'   data = adrs_f
#' )
#'
#' # Define groupings of BMRKR2 levels.
#' h_odds_ratio_subgroups_df(
#'   variables = list(
#'     rsp = "rsp",
#'     arm = "ARM",
#'     subgroups = c("SEX", "BMRKR2")
#'   ),
#'   data = adrs_f,
#'   groups_lists = list(
#'     BMRKR2 = list(
#'       "low" = "LOW",
#'       "low/medium" = c("LOW", "MEDIUM"),
#'       "low/medium/high" = c("LOW", "MEDIUM", "HIGH")
#'     )
#'   )
#' )
#'
#' @export
h_odds_ratio_subgroups_df <- function(variables,
                                      data,
                                      groups_lists = list(),
                                      conf_level = 0.95,
                                      method = NULL,
                                      label_all = "All Patients") {
  if ("strat" %in% names(variables)) {
    warning(
      "Warning: the `strat` element name of the `variables` list argument to `h_odds_ratio_subgroups_df() ",
      "was deprecated in tern 0.9.4.\n  ",
      "Please use the name `strata` instead of `strat` in the `variables` argument."
    )
    variables[["strata"]] <- variables[["strat"]]
  }

  checkmate::assert_character(variables$rsp)
  checkmate::assert_character(variables$arm)
  checkmate::assert_character(variables$subgroups, null.ok = TRUE)
  checkmate::assert_character(variables$strata, null.ok = TRUE)
  assert_df_with_factors(data, list(val = variables$arm), min.levels = 2, max.levels = 2)
  assert_df_with_variables(data, variables)
  checkmate::assert_string(label_all)

  strata_data <- if (is.null(variables$strata)) {
    NULL
  } else {
    data[, variables$strata, drop = FALSE]
  }

  # Add All Patients.
  result_all <- h_odds_ratio_df(
    rsp = data[[variables$rsp]],
    arm = data[[variables$arm]],
    strata_data = strata_data,
    conf_level = conf_level,
    method = method
  )
  result_all$subgroup <- label_all
  result_all$var <- "ALL"
  result_all$var_label <- label_all
  result_all$row_type <- "content"

  if (is.null(variables$subgroups)) {
    result_all
  } else {
    l_data <- h_split_by_subgroups(data, variables$subgroups, groups_lists = groups_lists)

    l_result <- lapply(l_data, function(grp) {
      grp_strata_data <- if (is.null(variables$strata)) {
        NULL
      } else {
        grp$df[, variables$strata, drop = FALSE]
      }

      result <- h_odds_ratio_df(
        rsp = grp$df[[variables$rsp]],
        arm = grp$df[[variables$arm]],
        strata_data = grp_strata_data,
        conf_level = conf_level,
        method = method
      )
      result_labels <- grp$df_labels[rep(1, times = nrow(result)), ]
      cbind(result, result_labels)
    })

    result_subgroups <- do.call(rbind, args = c(l_result, make.row.names = FALSE))
    result_subgroups$row_type <- "analysis"

    rbind(
      result_all,
      result_subgroups
    )
  }
}

#' Tabulate biomarker effects on survival by subgroup
#'
#' @description `r lifecycle::badge("stable")`
#'
#' The [tabulate_survival_biomarkers()] function creates a layout element to tabulate the estimated effects of multiple
#' continuous biomarker variables on survival across subgroups, returning statistics including median survival time and
#' hazard ratio for each population subgroup. The table is created from `df`, a list of data frames returned by
#' [extract_survival_biomarkers()], with the statistics to include specified via the `vars` parameter.
#'
#' A forest plot can be created from the resulting table using the [g_forest()] function.
#'
#' @inheritParams fit_coxreg_multivar
#' @inheritParams survival_duration_subgroups
#' @inheritParams argument_convention
#' @param df (`data.frame`)\cr containing all analysis variables, as returned by
#'   [extract_survival_biomarkers()].
#' @param vars (`character`)\cr the names of statistics to be reported among:
#'   * `n_tot_events`: Total number of events per group.
#'   * `n_tot`: Total number of observations per group.
#'   * `median`: Median survival time.
#'   * `hr`: Hazard ratio.
#'   * `ci`: Confidence interval of hazard ratio.
#'   * `pval`: p-value of the effect.
#'   Note, one of the statistics `n_tot` and `n_tot_events`, as well as both `hr` and `ci` are required.
#'
#' @details These functions create a layout starting from a data frame which contains
#'   the required statistics. The tables are then typically used as input for forest plots.
#'
#' @examples
#' library(dplyr)
#'
#' adtte <- tern_ex_adtte
#'
#' # Save variable labels before data processing steps.
#' adtte_labels <- formatters::var_labels(adtte)
#'
#' adtte_f <- adtte %>%
#'   filter(PARAMCD == "OS") %>%
#'   mutate(
#'     AVALU = as.character(AVALU),
#'     is_event = CNSR == 0
#'   )
#' labels <- c("AVALU" = adtte_labels[["AVALU"]], "is_event" = "Event Flag")
#' formatters::var_labels(adtte_f)[names(labels)] <- labels
#'
#' # Typical analysis of two continuous biomarkers `BMRKR1` and `AGE`,
#' # in multiple regression models containing one covariate `RACE`,
#' # as well as one stratification variable `STRATA1`. The subgroups
#' # are defined by the levels of `BMRKR2`.
#'
#' df <- extract_survival_biomarkers(
#'   variables = list(
#'     tte = "AVAL",
#'     is_event = "is_event",
#'     biomarkers = c("BMRKR1", "AGE"),
#'     strata = "STRATA1",
#'     covariates = "SEX",
#'     subgroups = "BMRKR2"
#'   ),
#'   label_all = "Total Patients",
#'   data = adtte_f
#' )
#' df
#'
#' # Here we group the levels of `BMRKR2` manually.
#' df_grouped <- extract_survival_biomarkers(
#'   variables = list(
#'     tte = "AVAL",
#'     is_event = "is_event",
#'     biomarkers = c("BMRKR1", "AGE"),
#'     strata = "STRATA1",
#'     covariates = "SEX",
#'     subgroups = "BMRKR2"
#'   ),
#'   data = adtte_f,
#'   groups_lists = list(
#'     BMRKR2 = list(
#'       "low" = "LOW",
#'       "low/medium" = c("LOW", "MEDIUM"),
#'       "low/medium/high" = c("LOW", "MEDIUM", "HIGH")
#'     )
#'   )
#' )
#' df_grouped
#'
#' @name survival_biomarkers_subgroups
#' @order 1
NULL

#' Prepare survival data estimates for multiple biomarkers in a single data frame
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Prepares estimates for number of events, patients and median survival times, as well as hazard ratio estimates,
#' confidence intervals and p-values, for multiple biomarkers across population subgroups in a single data frame.
#' `variables` corresponds to the names of variables found in `data`, passed as a named `list` and requires elements
#' `tte`, `is_event`, `biomarkers` (vector of continuous biomarker variables), and optionally `subgroups` and `strata`.
#' `groups_lists` optionally specifies groupings for `subgroups` variables.
#'
#' @inheritParams argument_convention
#' @inheritParams fit_coxreg_multivar
#' @inheritParams survival_duration_subgroups
#'
#' @return A `data.frame` with columns `biomarker`, `biomarker_label`, `n_tot`, `n_tot_events`,
#'   `median`, `hr`, `lcl`, `ucl`, `conf_level`, `pval`, `pval_label`, `subgroup`, `var`,
#'   `var_label`, and `row_type`.
#'
#' @seealso [h_coxreg_mult_cont_df()] which is used internally, [tabulate_survival_biomarkers()].
#'
#' @export
extract_survival_biomarkers <- function(variables,
                                        data,
                                        groups_lists = list(),
                                        control = control_coxreg(),
                                        label_all = "All Patients") {
  if ("strat" %in% names(variables)) {
    warning(
      "Warning: the `strat` element name of the `variables` list argument to `extract_survival_biomarkers() ",
      "was deprecated in tern 0.9.4.\n  ",
      "Please use the name `strata` instead of `strat` in the `variables` argument."
    )
    variables[["strata"]] <- variables[["strat"]]
  }

  checkmate::assert_list(variables)
  checkmate::assert_character(variables$subgroups, null.ok = TRUE)
  checkmate::assert_string(label_all)

  # Start with all patients.
  result_all <- h_coxreg_mult_cont_df(
    variables = variables,
    data = data,
    control = control
  )
  result_all$subgroup <- label_all
  result_all$var <- "ALL"
  result_all$var_label <- label_all
  result_all$row_type <- "content"
  if (is.null(variables$subgroups)) {
    # Only return result for all patients.
    result_all
  } else {
    # Add subgroups results.
    l_data <- h_split_by_subgroups(
      data,
      variables$subgroups,
      groups_lists = groups_lists
    )
    l_result <- lapply(l_data, function(grp) {
      result <- h_coxreg_mult_cont_df(
        variables = variables,
        data = grp$df,
        control = control
      )
      result_labels <- grp$df_labels[rep(1, times = nrow(result)), ]
      cbind(result, result_labels)
    })
    result_subgroups <- do.call(rbind, args = c(l_result, make.row.names = FALSE))
    result_subgroups$row_type <- "analysis"
    rbind(
      result_all,
      result_subgroups
    )
  }
}

#' @describeIn survival_biomarkers_subgroups Table-creating function which creates a table
#'   summarizing biomarker effects on survival by subgroup.
#'
#' @param label_all `r lifecycle::badge("deprecated")`\cr please assign the `label_all` parameter within the
#'   [extract_survival_biomarkers()] function when creating `df`.
#'
#' @return An `rtables` table summarizing biomarker effects on survival by subgroup.
#'
#' @note In contrast to [tabulate_survival_subgroups()] this tabulation function does
#'   not start from an input layout `lyt`. This is because internally the table is
#'   created by combining multiple subtables.
#'
#' @seealso [h_tab_surv_one_biomarker()] which is used internally, [extract_survival_biomarkers()].
#'
#' @examples
#' ## Table with default columns.
#' tabulate_survival_biomarkers(df)
#'
#' ## Table with a manually chosen set of columns: leave out "pval", reorder.
#' tab <- tabulate_survival_biomarkers(
#'   df = df,
#'   vars = c("n_tot_events", "ci", "n_tot", "median", "hr"),
#'   time_unit = as.character(adtte_f$AVALU[1])
#' )
#'
#' ## Finally produce the forest plot.
#' \donttest{
#' g_forest(tab, xlim = c(0.8, 1.2))
#' }
#'
#' @export
#' @order 2
tabulate_survival_biomarkers <- function(df,
                                         vars = c("n_tot", "n_tot_events", "median", "hr", "ci", "pval"),
                                         groups_lists = list(),
                                         control = control_coxreg(),
                                         label_all = lifecycle::deprecated(),
                                         time_unit = NULL,
                                         na_str = default_na_str(),
                                         .indent_mods = 0L) {
  if (lifecycle::is_present(label_all)) {
    lifecycle::deprecate_warn(
      "0.9.5", "tabulate_survival_biomarkers(label_all)",
      details = paste(
        "Please assign the `label_all` parameter within the",
        "`extract_survival_biomarkers()` function when creating `df`."
      )
    )
  }

  checkmate::assert_data_frame(df)
  checkmate::assert_character(df$biomarker)
  checkmate::assert_character(df$biomarker_label)
  checkmate::assert_subset(vars, get_stats("tabulate_survival_biomarkers"))

  extra_args <- list(groups_lists = groups_lists, control = control)

  df_subs <- split(df, f = df$biomarker)
  tabs <- lapply(df_subs, FUN = function(df_sub) {
    tab_sub <- h_tab_surv_one_biomarker(
      df = df_sub,
      vars = vars,
      time_unit = time_unit,
      na_str = na_str,
      .indent_mods = .indent_mods,
      extra_args = extra_args
    )
    # Insert label row as first row in table.
    label_at_path(tab_sub, path = row_paths(tab_sub)[[1]][1]) <- df_sub$biomarker_label[1]
    tab_sub
  })
  result <- do.call(rbind, tabs)

  n_tot_ids <- grep("^n_tot", vars)
  hr_id <- match("hr", vars)
  ci_id <- match("ci", vars)
  structure(
    result,
    forest_header = paste0(c("Higher", "Lower"), "\nBetter"),
    col_x = hr_id,
    col_ci = ci_id,
    col_symbol_size = n_tot_ids[1]
  )
}

#' Proportion difference estimation
#'
#' @description `r lifecycle::badge("stable")`
#'
#' The analysis function [estimate_proportion_diff()] creates a layout element to estimate the difference in proportion
#' of responders within a studied population. The primary analysis variable, `vars`, is a logical variable indicating
#' whether a response has occurred for each record. See the `method` parameter for options of methods to use when
#' constructing the confidence interval of the proportion difference. A stratification variable can be supplied via the
#' `strata` element of the `variables` argument.
#'
#'
#' @inheritParams prop_diff_strat_nc
#' @inheritParams argument_convention
#' @param method (`string`)\cr the method used for the confidence interval estimation.
#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("estimate_proportion_diff")`
#'   to see available statistics for this function.
#'
#' @seealso [d_proportion_diff()]
#'
#' @name prop_diff
#' @order 1
NULL

#' @describeIn prop_diff Statistics function estimating the difference
#'   in terms of responder proportion.
#'
#' @return
#' * `s_proportion_diff()` returns a named list of elements `diff` and `diff_ci`.
#'
#' @note When performing an unstratified analysis, methods `"cmh"`, `"strat_newcombe"`, and `"strat_newcombecc"` are
#'   not permitted.
#'
#' @examples
#' s_proportion_diff(
#'   df = subset(dta, grp == "A"),
#'   .var = "rsp",
#'   .ref_group = subset(dta, grp == "B"),
#'   .in_ref_col = FALSE,
#'   conf_level = 0.90,
#'   method = "ha"
#' )
#'
#' # CMH example with strata
#' s_proportion_diff(
#'   df = subset(dta, grp == "A"),
#'   .var = "rsp",
#'   .ref_group = subset(dta, grp == "B"),
#'   .in_ref_col = FALSE,
#'   variables = list(strata = c("f1", "f2")),
#'   conf_level = 0.90,
#'   method = "cmh"
#' )
#'
#' @export
s_proportion_diff <- function(df,
                              .var,
                              .ref_group,
                              .in_ref_col,
                              variables = list(strata = NULL),
                              conf_level = 0.95,
                              method = c(
                                "waldcc", "wald", "cmh",
                                "ha", "newcombe", "newcombecc",
                                "strat_newcombe", "strat_newcombecc"
                              ),
                              weights_method = "cmh") {
  method <- match.arg(method)
  if (is.null(variables$strata) && checkmate::test_subset(method, c("cmh", "strat_newcombe", "strat_newcombecc"))) {
    stop(paste(
      "When performing an unstratified analysis, methods 'cmh', 'strat_newcombe', and 'strat_newcombecc' are not",
      "permitted. Please choose a different method."
    ))
  }
  y <- list(diff = "", diff_ci = "")

  if (!.in_ref_col) {
    rsp <- c(.ref_group[[.var]], df[[.var]])
    grp <- factor(
      rep(
        c("ref", "Not-ref"),
        c(nrow(.ref_group), nrow(df))
      ),
      levels = c("ref", "Not-ref")
    )

    if (!is.null(variables$strata)) {
      strata_colnames <- variables$strata
      checkmate::assert_character(strata_colnames, null.ok = FALSE)
      strata_vars <- stats::setNames(as.list(strata_colnames), strata_colnames)

      assert_df_with_variables(df, strata_vars)
      assert_df_with_variables(.ref_group, strata_vars)

      # Merging interaction strata for reference group rows data and remaining
      strata <- c(
        interaction(.ref_group[strata_colnames]),
        interaction(df[strata_colnames])
      )
      strata <- as.factor(strata)
    }

    # Defining the std way to calculate weights for strat_newcombe
    if (!is.null(variables$weights_method)) {
      weights_method <- variables$weights_method
    } else {
      weights_method <- "cmh"
    }

    y <- switch(method,
      "wald" = prop_diff_wald(rsp, grp, conf_level, correct = FALSE),
      "waldcc" = prop_diff_wald(rsp, grp, conf_level, correct = TRUE),
      "ha" = prop_diff_ha(rsp, grp, conf_level),
      "newcombe" = prop_diff_nc(rsp, grp, conf_level, correct = FALSE),
      "newcombecc" = prop_diff_nc(rsp, grp, conf_level, correct = TRUE),
      "strat_newcombe" = prop_diff_strat_nc(rsp,
        grp,
        strata,
        weights_method,
        conf_level,
        correct = FALSE
      ),
      "strat_newcombecc" = prop_diff_strat_nc(rsp,
        grp,
        strata,
        weights_method,
        conf_level,
        correct = TRUE
      ),
      "cmh" = prop_diff_cmh(rsp, grp, strata, conf_level)[c("diff", "diff_ci")]
    )

    y$diff <- y$diff * 100
    y$diff_ci <- y$diff_ci * 100
  }

  attr(y$diff, "label") <- "Difference in Response rate (%)"
  attr(y$diff_ci, "label") <- d_proportion_diff(
    conf_level, method,
    long = FALSE
  )

  y
}

#' @describeIn prop_diff Formatted analysis function which is used as `afun` in `estimate_proportion_diff()`.
#'
#' @return
#' * `a_proportion_diff()` returns the corresponding list with formatted [rtables::CellValue()].
#'
#' @examples
#' a_proportion_diff(
#'   df = subset(dta, grp == "A"),
#'   .var = "rsp",
#'   .ref_group = subset(dta, grp == "B"),
#'   .in_ref_col = FALSE,
#'   conf_level = 0.90,
#'   method = "ha"
#' )
#'
#' @export
a_proportion_diff <- make_afun(
  s_proportion_diff,
  .formats = c(diff = "xx.x", diff_ci = "(xx.x, xx.x)"),
  .indent_mods = c(diff = 0L, diff_ci = 1L)
)

#' @describeIn prop_diff Layout-creating function which can take statistics function arguments
#'   and additional format arguments. This function is a wrapper for [rtables::analyze()].
#'
#' @return
#' * `estimate_proportion_diff()` returns a layout object suitable for passing to further layouting functions,
#'   or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
#'   the statistics from `s_proportion_diff()` to the table layout.
#'
#' @examples
#' ## "Mid" case: 4/4 respond in group A, 1/2 respond in group B.
#' nex <- 100 # Number of example rows
#' dta <- data.frame(
#'   "rsp" = sample(c(TRUE, FALSE), nex, TRUE),
#'   "grp" = sample(c("A", "B"), nex, TRUE),
#'   "f1" = sample(c("a1", "a2"), nex, TRUE),
#'   "f2" = sample(c("x", "y", "z"), nex, TRUE),
#'   stringsAsFactors = TRUE
#' )
#'
#' l <- basic_table() %>%
#'   split_cols_by(var = "grp", ref_group = "B") %>%
#'   estimate_proportion_diff(
#'     vars = "rsp",
#'     conf_level = 0.90,
#'     method = "ha"
#'   )
#'
#' build_table(l, df = dta)
#'
#' @export
#' @order 2
estimate_proportion_diff <- function(lyt,
                                     vars,
                                     variables = list(strata = NULL),
                                     conf_level = 0.95,
                                     method = c(
                                       "waldcc", "wald", "cmh",
                                       "ha", "newcombe", "newcombecc",
                                       "strat_newcombe", "strat_newcombecc"
                                     ),
                                     weights_method = "cmh",
                                     na_str = default_na_str(),
                                     nested = TRUE,
                                     ...,
                                     var_labels = vars,
                                     show_labels = "hidden",
                                     table_names = vars,
                                     .stats = NULL,
                                     .formats = NULL,
                                     .labels = NULL,
                                     .indent_mods = NULL) {
  extra_args <- list(
    variables = variables, conf_level = conf_level, method = method, weights_method = weights_method, ...
  )

  afun <- make_afun(
    a_proportion_diff,
    .stats = .stats,
    .formats = .formats,
    .labels = .labels,
    .indent_mods = .indent_mods
  )

  analyze(
    lyt,
    vars,
    afun = afun,
    var_labels = var_labels,
    na_str = na_str,
    nested = nested,
    extra_args = extra_args,
    show_labels = show_labels,
    table_names = table_names
  )
}

#' Check proportion difference arguments
#'
#' Verifies that and/or convert arguments into valid values to be used in the
#' estimation of difference in responder proportions.
#'
#' @inheritParams prop_diff
#' @inheritParams prop_diff_wald
#'
#' @keywords internal
check_diff_prop_ci <- function(rsp,
                               grp,
                               strata = NULL,
                               conf_level,
                               correct = NULL) {
  checkmate::assert_logical(rsp, any.missing = FALSE)
  checkmate::assert_factor(grp, len = length(rsp), any.missing = FALSE, n.levels = 2)
  checkmate::assert_number(conf_level, lower = 0, upper = 1)
  checkmate::assert_flag(correct, null.ok = TRUE)

  if (!is.null(strata)) {
    checkmate::assert_factor(strata, len = length(rsp))
  }

  invisible()
}

#' Description of method used for proportion comparison
#'
#' @description `r lifecycle::badge("stable")`
#'
#' This is an auxiliary function that describes the analysis in
#' [s_proportion_diff()].
#'
#' @inheritParams s_proportion_diff
#' @param long (`flag`)\cr whether a long (`TRUE`) or a short (`FALSE`, default) description is required.
#'
#' @return A `string` describing the analysis.
#'
#' @seealso [prop_diff]
#'
#' @export
d_proportion_diff <- function(conf_level,
                              method,
                              long = FALSE) {
  label <- paste0(conf_level * 100, "% CI")
  if (long) {
    label <- paste(
      label,
      ifelse(
        method == "cmh",
        "for adjusted difference",
        "for difference"
      )
    )
  }

  method_part <- switch(method,
    "cmh" = "CMH, without correction",
    "waldcc" = "Wald, with correction",
    "wald" = "Wald, without correction",
    "ha" = "Anderson-Hauck",
    "newcombe" = "Newcombe, without correction",
    "newcombecc" = "Newcombe, with correction",
    "strat_newcombe" = "Stratified Newcombe, without correction",
    "strat_newcombecc" = "Stratified Newcombe, with correction",
    stop(paste(method, "does not have a description"))
  )
  paste0(label, " (", method_part, ")")
}

#' Helper functions to calculate proportion difference
#'
#' @description `r lifecycle::badge("stable")`
#'
#' @inheritParams argument_convention
#' @inheritParams prop_diff
#' @param grp (`factor`)\cr vector assigning observations to one out of two groups
#'   (e.g. reference and treatment group).
#'
#' @return A named `list` of elements `diff` (proportion difference) and `diff_ci`
#'   (proportion difference confidence interval).
#'
#' @seealso [prop_diff()] for implementation of these helper functions.
#'
#' @name h_prop_diff
NULL

#' @describeIn h_prop_diff The Wald interval follows the usual textbook
#'   definition for a single proportion confidence interval using the normal
#'   approximation. It is possible to include a continuity correction for Wald's
#'   interval.
#'
#' @param correct (`flag`)\cr whether to include the continuity correction. For further
#'   information, see [stats::prop.test()].
#'
#' @examples
#' # Wald confidence interval
#' set.seed(2)
#' rsp <- sample(c(TRUE, FALSE), replace = TRUE, size = 20)
#' grp <- factor(c(rep("A", 10), rep("B", 10)))
#'
#' prop_diff_wald(rsp = rsp, grp = grp, conf_level = 0.95, correct = FALSE)
#'
#' @export
prop_diff_wald <- function(rsp,
                           grp,
                           conf_level = 0.95,
                           correct = FALSE) {
  if (isTRUE(correct)) {
    mthd <- "waldcc"
  } else {
    mthd <- "wald"
  }
  grp <- as_factor_keep_attributes(grp)
  check_diff_prop_ci(
    rsp = rsp, grp = grp, conf_level = conf_level, correct = correct
  )

  # check if binary response is coded as logical
  checkmate::assert_logical(rsp, any.missing = FALSE)
  checkmate::assert_factor(grp, len = length(rsp), any.missing = FALSE, n.levels = 2)

  tbl <- table(grp, factor(rsp, levels = c(TRUE, FALSE)))
  # x1 and n1 are non-reference groups.
  diff_ci <- desctools_binom(
    x1 = tbl[2], n1 = sum(tbl[2], tbl[4]),
    x2 = tbl[1], n2 = sum(tbl[1], tbl[3]),
    conf.level = conf_level,
    method = mthd
  )

  list(
    "diff" = unname(diff_ci[, "est"]),
    "diff_ci" = unname(diff_ci[, c("lwr.ci", "upr.ci")])
  )
}

#' @describeIn h_prop_diff Anderson-Hauck confidence interval.
#'
#' @examples
#' # Anderson-Hauck confidence interval
#' ## "Mid" case: 3/4 respond in group A, 1/2 respond in group B.
#' rsp <- c(TRUE, FALSE, FALSE, TRUE, TRUE, TRUE)
#' grp <- factor(c("A", "B", "A", "B", "A", "A"), levels = c("B", "A"))
#'
#' prop_diff_ha(rsp = rsp, grp = grp, conf_level = 0.90)
#'
#' ## Edge case: Same proportion of response in A and B.
#' rsp <- c(TRUE, FALSE, TRUE, FALSE)
#' grp <- factor(c("A", "A", "B", "B"), levels = c("A", "B"))
#'
#' prop_diff_ha(rsp = rsp, grp = grp, conf_level = 0.6)
#'
#' @export
prop_diff_ha <- function(rsp,
                         grp,
                         conf_level) {
  grp <- as_factor_keep_attributes(grp)
  check_diff_prop_ci(rsp = rsp, grp = grp, conf_level = conf_level)

  tbl <- table(grp, factor(rsp, levels = c(TRUE, FALSE)))
  # x1 and n1 are non-reference groups.
  ci <- desctools_binom(
    x1 = tbl[2], n1 = sum(tbl[2], tbl[4]),
    x2 = tbl[1], n2 = sum(tbl[1], tbl[3]),
    conf.level = conf_level,
    method = "ha"
  )
  list(
    "diff" = unname(ci[, "est"]),
    "diff_ci" = unname(ci[, c("lwr.ci", "upr.ci")])
  )
}

#' @describeIn h_prop_diff Newcombe confidence interval. It is based on
#'   the Wilson score confidence interval for a single binomial proportion.
#'
#' @examples
#' # Newcombe confidence interval
#'
#' set.seed(1)
#' rsp <- c(
#'   sample(c(TRUE, FALSE), size = 40, prob = c(3 / 4, 1 / 4), replace = TRUE),
#'   sample(c(TRUE, FALSE), size = 40, prob = c(1 / 2, 1 / 2), replace = TRUE)
#' )
#' grp <- factor(rep(c("A", "B"), each = 40), levels = c("B", "A"))
#' table(rsp, grp)
#'
#' prop_diff_nc(rsp = rsp, grp = grp, conf_level = 0.9)
#'
#' @export
prop_diff_nc <- function(rsp,
                         grp,
                         conf_level,
                         correct = FALSE) {
  if (isTRUE(correct)) {
    mthd <- "scorecc"
  } else {
    mthd <- "score"
  }
  grp <- as_factor_keep_attributes(grp)
  check_diff_prop_ci(rsp = rsp, grp = grp, conf_level = conf_level)

  p_grp <- tapply(rsp, grp, mean)
  diff_p <- unname(diff(p_grp))
  tbl <- table(grp, factor(rsp, levels = c(TRUE, FALSE)))
  ci <- desctools_binom(
    # x1 and n1 are non-reference groups.
    x1 = tbl[2], n1 = sum(tbl[2], tbl[4]),
    x2 = tbl[1], n2 = sum(tbl[1], tbl[3]),
    conf.level = conf_level,
    method = mthd
  )
  list(
    "diff" = unname(ci[, "est"]),
    "diff_ci" = unname(ci[, c("lwr.ci", "upr.ci")])
  )
}

#' @describeIn h_prop_diff Calculates the weighted difference. This is defined as the difference in
#'   response rates between the experimental treatment group and the control treatment group, adjusted
#'   for stratification factors by applying Cochran-Mantel-Haenszel (CMH) weights. For the CMH chi-squared
#'   test, use [stats::mantelhaen.test()].
#'
#' @param strata (`factor`)\cr variable with one level per stratum and same length as `rsp`.
#'
#' @examples
#' # Cochran-Mantel-Haenszel confidence interval
#'
#' set.seed(2)
#' rsp <- sample(c(TRUE, FALSE), 100, TRUE)
#' grp <- sample(c("Placebo", "Treatment"), 100, TRUE)
#' grp <- factor(grp, levels = c("Placebo", "Treatment"))
#' strata_data <- data.frame(
#'   "f1" = sample(c("a", "b"), 100, TRUE),
#'   "f2" = sample(c("x", "y", "z"), 100, TRUE),
#'   stringsAsFactors = TRUE
#' )
#'
#' prop_diff_cmh(
#'   rsp = rsp, grp = grp, strata = interaction(strata_data),
#'   conf_level = 0.90
#' )
#'
#' @export
prop_diff_cmh <- function(rsp,
                          grp,
                          strata,
                          conf_level = 0.95) {
  grp <- as_factor_keep_attributes(grp)
  strata <- as_factor_keep_attributes(strata)
  check_diff_prop_ci(
    rsp = rsp, grp = grp, conf_level = conf_level, strata = strata
  )

  if (any(tapply(rsp, strata, length) < 5)) {
    warning("Less than 5 observations in some strata.")
  }

  # first dimension: FALSE, TRUE
  # 2nd dimension: CONTROL, TX
  # 3rd dimension: levels of strata
  # rsp as factor rsp to handle edge case of no FALSE (or TRUE) rsp records
  t_tbl <- table(
    factor(rsp, levels = c("FALSE", "TRUE")),
    grp,
    strata
  )
  n1 <- colSums(t_tbl[1:2, 1, ])
  n2 <- colSums(t_tbl[1:2, 2, ])
  p1 <- t_tbl[2, 1, ] / n1
  p2 <- t_tbl[2, 2, ] / n2
  # CMH weights
  use_stratum <- (n1 > 0) & (n2 > 0)
  n1 <- n1[use_stratum]
  n2 <- n2[use_stratum]
  p1 <- p1[use_stratum]
  p2 <- p2[use_stratum]
  wt <- (n1 * n2 / (n1 + n2))
  wt_normalized <- wt / sum(wt)
  est1 <- sum(wt_normalized * p1)
  est2 <- sum(wt_normalized * p2)
  estimate <- c(est1, est2)
  names(estimate) <- levels(grp)
  se1 <- sqrt(sum(wt_normalized^2 * p1 * (1 - p1) / n1))
  se2 <- sqrt(sum(wt_normalized^2 * p2 * (1 - p2) / n2))
  z <- stats::qnorm((1 + conf_level) / 2)
  err1 <- z * se1
  err2 <- z * se2
  ci1 <- c((est1 - err1), (est1 + err1))
  ci2 <- c((est2 - err2), (est2 + err2))
  estimate_ci <- list(ci1, ci2)
  names(estimate_ci) <- levels(grp)
  diff_est <- est2 - est1
  se_diff <- sqrt(sum(((p1 * (1 - p1) / n1) + (p2 * (1 - p2) / n2)) * wt_normalized^2))
  diff_ci <- c(diff_est - z * se_diff, diff_est + z * se_diff)

  list(
    prop = estimate,
    prop_ci = estimate_ci,
    diff = diff_est,
    diff_ci = diff_ci,
    weights = wt_normalized,
    n1 = n1,
    n2 = n2
  )
}

#' @describeIn h_prop_diff Calculates the stratified Newcombe confidence interval and difference in response
#'   rates between the experimental treatment group and the control treatment group, adjusted for stratification
#'   factors. This implementation follows closely the one proposed by \insertCite{Yan2010-jt;textual}{tern}.
#'   Weights can be estimated from the heuristic proposed in [prop_strat_wilson()] or from CMH-derived weights
#'   (see [prop_diff_cmh()]).
#'
#' @param strata (`factor`)\cr variable with one level per stratum and same length as `rsp`.
#' @param weights_method (`string`)\cr weights method. Can be either `"cmh"` or `"heuristic"`
#'   and directs the way weights are estimated.
#'
#' @references
#' \insertRef{Yan2010-jt}{tern}
#'
#' @examples
#' # Stratified Newcombe confidence interval
#'
#' set.seed(2)
#' data_set <- data.frame(
#'   "rsp" = sample(c(TRUE, FALSE), 100, TRUE),
#'   "f1" = sample(c("a", "b"), 100, TRUE),
#'   "f2" = sample(c("x", "y", "z"), 100, TRUE),
#'   "grp" = sample(c("Placebo", "Treatment"), 100, TRUE),
#'   stringsAsFactors = TRUE
#' )
#'
#' prop_diff_strat_nc(
#'   rsp = data_set$rsp, grp = data_set$grp, strata = interaction(data_set[2:3]),
#'   weights_method = "cmh",
#'   conf_level = 0.90
#' )
#'
#' prop_diff_strat_nc(
#'   rsp = data_set$rsp, grp = data_set$grp, strata = interaction(data_set[2:3]),
#'   weights_method = "wilson_h",
#'   conf_level = 0.90
#' )
#'
#' @export
prop_diff_strat_nc <- function(rsp,
                               grp,
                               strata,
                               weights_method = c("cmh", "wilson_h"),
                               conf_level = 0.95,
                               correct = FALSE) {
  weights_method <- match.arg(weights_method)
  grp <- as_factor_keep_attributes(grp)
  strata <- as_factor_keep_attributes(strata)
  check_diff_prop_ci(
    rsp = rsp, grp = grp, conf_level = conf_level, strata = strata
  )
  checkmate::assert_number(conf_level, lower = 0, upper = 1)
  checkmate::assert_flag(correct)
  if (any(tapply(rsp, strata, length) < 5)) {
    warning("Less than 5 observations in some strata.")
  }

  rsp_by_grp <- split(rsp, f = grp)
  strata_by_grp <- split(strata, f = grp)

  # Finding the weights
  weights <- if (identical(weights_method, "cmh")) {
    prop_diff_cmh(rsp = rsp, grp = grp, strata = strata)$weights
  } else if (identical(weights_method, "wilson_h")) {
    prop_strat_wilson(rsp, strata, conf_level = conf_level, correct = correct)$weights
  }
  weights[levels(strata)[!levels(strata) %in% names(weights)]] <- 0

  # Calculating lower (`l`) and upper (`u`) confidence bounds per group.
  strat_wilson_by_grp <- Map(
    prop_strat_wilson,
    rsp = rsp_by_grp,
    strata = strata_by_grp,
    weights = list(weights, weights),
    conf_level = conf_level,
    correct = correct
  )

  ci_ref <- strat_wilson_by_grp[[1]]
  ci_trt <- strat_wilson_by_grp[[2]]
  l_ref <- as.numeric(ci_ref$conf_int[1])
  u_ref <- as.numeric(ci_ref$conf_int[2])
  l_trt <- as.numeric(ci_trt$conf_int[1])
  u_trt <- as.numeric(ci_trt$conf_int[2])

  # Estimating the diff and n_ref, n_trt (it allows different weights to be used)
  t_tbl <- table(
    factor(rsp, levels = c("FALSE", "TRUE")),
    grp,
    strata
  )
  n_ref <- colSums(t_tbl[1:2, 1, ])
  n_trt <- colSums(t_tbl[1:2, 2, ])
  use_stratum <- (n_ref > 0) & (n_trt > 0)
  n_ref <- n_ref[use_stratum]
  n_trt <- n_trt[use_stratum]
  p_ref <- t_tbl[2, 1, use_stratum] / n_ref
  p_trt <- t_tbl[2, 2, use_stratum] / n_trt
  est1 <- sum(weights * p_ref)
  est2 <- sum(weights * p_trt)
  diff_est <- est2 - est1

  lambda1 <- sum(weights^2 / n_ref)
  lambda2 <- sum(weights^2 / n_trt)
  z <- stats::qnorm((1 + conf_level) / 2)

  lower <- diff_est - z * sqrt(lambda2 * l_trt * (1 - l_trt) + lambda1 * u_ref * (1 - u_ref))
  upper <- diff_est + z * sqrt(lambda1 * l_ref * (1 - l_ref) + lambda2 * u_trt * (1 - u_trt))

  list(
    "diff" = diff_est,
    "diff_ci" = c("lower" = lower, "upper" = upper)
  )
}

#' Count number of patients and sum exposure across all patients in columns
#'
#' @description `r lifecycle::badge("stable")`
#'
#' The analyze function [analyze_patients_exposure_in_cols()] creates a layout element to count total numbers of
#' patients and sum an analysis value (i.e. exposure) across all patients in columns.
#'
#' The primary analysis variable `ex_var` is the exposure variable used to calculate the `sum_exposure` statistic. The
#' `id` variable is used to uniquely identify patients in the data such that only unique patients are counted in the
#' `n_patients` statistic, and the `var` variable is used to create a row split if needed. The percentage returned as
#' part of the `n_patients` statistic is the proportion of all records that correspond to a unique patient.
#'
#' The summarize function [summarize_patients_exposure_in_cols()] performs the same function as
#' [analyze_patients_exposure_in_cols()] except it creates content rows, not data rows, to summarize the current table
#' row/column context and operates on the level of the latest row split or the root of the table if no row splits have
#' occurred.
#'
#' If a column split has not yet been performed in the table, `col_split` must be set to `TRUE` for the first call of
#' [analyze_patients_exposure_in_cols()] or [summarize_patients_exposure_in_cols()].
#'
#' @inheritParams argument_convention
#' @param ex_var (`string`)\cr name of the variable in `df` containing exposure values.
#' @param custom_label (`string` or `NULL`)\cr if provided and `labelstr` is empty, this will be used as label.
#' @param .stats (`character`)\cr statistics to select for the table. Run
#'   `get_stats("analyze_patients_exposure_in_cols")` to see available statistics for this function.
#'
#' @name summarize_patients_exposure_in_cols
#' @order 1
NULL

#' @describeIn summarize_patients_exposure_in_cols Statistics function which counts numbers
#'   of patients and the sum of exposure across all patients.
#'
#' @return
#' * `s_count_patients_sum_exposure()` returns a named `list` with the statistics:
#'   * `n_patients`: Number of unique patients in `df`.
#'   * `sum_exposure`: Sum of `ex_var` across all patients in `df`.
#'
#' @keywords internal
s_count_patients_sum_exposure <- function(df,
                                          ex_var = "AVAL",
                                          id = "USUBJID",
                                          labelstr = "",
                                          .stats = c("n_patients", "sum_exposure"),
                                          .N_col, # nolint
                                          custom_label = NULL) {
  assert_df_with_variables(df, list(ex_var = ex_var, id = id))
  checkmate::assert_string(id)
  checkmate::assert_string(labelstr)
  checkmate::assert_string(custom_label, null.ok = TRUE)
  checkmate::assert_numeric(df[[ex_var]])
  checkmate::assert_true(all(.stats %in% c("n_patients", "sum_exposure")))

  row_label <- if (labelstr != "") {
    labelstr
  } else if (!is.null(custom_label)) {
    custom_label
  } else {
    "Total patients numbers/person time"
  }

  y <- list()

  if ("n_patients" %in% .stats) {
    y$n_patients <-
      formatters::with_label(
        s_num_patients_content(
          df = df,
          .N_col = .N_col, # nolint
          .var = id,
          labelstr = ""
        )$unique,
        row_label
      )
  }
  if ("sum_exposure" %in% .stats) {
    y$sum_exposure <- formatters::with_label(sum(df[[ex_var]]), row_label)
  }
  y
}

#' @describeIn summarize_patients_exposure_in_cols Analysis function which is used as `afun` in
#'   [rtables::analyze_colvars()] within `analyze_patients_exposure_in_cols()` and as `cfun` in
#'   [rtables::summarize_row_groups()] within `summarize_patients_exposure_in_cols()`.
#'
#' @return
#' * `a_count_patients_sum_exposure()` returns formatted [rtables::CellValue()].
#'
#' @examples
#' a_count_patients_sum_exposure(
#'   df = df,
#'   var = "SEX",
#'   .N_col = nrow(df),
#'   .stats = "n_patients"
#' )
#'
#' @export
a_count_patients_sum_exposure <- function(df,
                                          var = NULL,
                                          ex_var = "AVAL",
                                          id = "USUBJID",
                                          add_total_level = FALSE,
                                          custom_label = NULL,
                                          labelstr = "",
                                          .N_col, # nolint
                                          .stats,
                                          .formats = list(n_patients = "xx (xx.x%)", sum_exposure = "xx")) {
  checkmate::assert_flag(add_total_level)

  if (!is.null(var)) {
    assert_df_with_variables(df, list(var = var))
    df[[var]] <- as.factor(df[[var]])
  }

  y <- list()
  if (is.null(var)) {
    y[[.stats]] <- list(Total = s_count_patients_sum_exposure(
      df = df,
      ex_var = ex_var,
      id = id,
      labelstr = labelstr,
      .N_col = .N_col,
      .stats = .stats,
      custom_label = custom_label
    )[[.stats]])
  } else {
    for (lvl in levels(df[[var]])) {
      y[[.stats]][[lvl]] <- s_count_patients_sum_exposure(
        df = subset(df, get(var) == lvl),
        ex_var = ex_var,
        id = id,
        labelstr = labelstr,
        .N_col = .N_col,
        .stats = .stats,
        custom_label = lvl
      )[[.stats]]
    }
    if (add_total_level) {
      y[[.stats]][["Total"]] <- s_count_patients_sum_exposure(
        df = df,
        ex_var = ex_var,
        id = id,
        labelstr = labelstr,
        .N_col = .N_col,
        .stats = .stats,
        custom_label = custom_label
      )[[.stats]]
    }
  }

  in_rows(.list = y[[.stats]], .formats = .formats[[.stats]])
}

#' @describeIn summarize_patients_exposure_in_cols Layout-creating function which can take statistics
#'   function arguments and additional format arguments. This function is a wrapper for
#'   [rtables::split_cols_by_multivar()] and [rtables::summarize_row_groups()].
#'
#' @return
#' * `summarize_patients_exposure_in_cols()` returns a layout object suitable for passing to further
#'   layouting functions, or to [rtables::build_table()]. Adding this function to an `rtable` layout will
#'   add formatted content rows, with the statistics from `s_count_patients_sum_exposure()` arranged in
#'   columns, to the table layout.
#'
#' @examples
#' lyt5 <- basic_table() %>%
#'   summarize_patients_exposure_in_cols(var = "AVAL", col_split = TRUE)
#'
#' result5 <- build_table(lyt5, df = df, alt_counts_df = adsl)
#' result5
#'
#' lyt6 <- basic_table() %>%
#'   summarize_patients_exposure_in_cols(var = "AVAL", col_split = TRUE, .stats = "sum_exposure")
#'
#' result6 <- build_table(lyt6, df = df, alt_counts_df = adsl)
#' result6
#'
#' @export
#' @order 3
summarize_patients_exposure_in_cols <- function(lyt, # nolint
                                                var,
                                                ex_var = "AVAL",
                                                id = "USUBJID",
                                                add_total_level = FALSE,
                                                custom_label = NULL,
                                                col_split = TRUE,
                                                na_str = default_na_str(),
                                                ...,
                                                .stats = c("n_patients", "sum_exposure"),
                                                .labels = c(n_patients = "Patients", sum_exposure = "Person time"),
                                                .indent_mods = NULL) {
  extra_args <- list(ex_var = ex_var, id = id, add_total_level = add_total_level, custom_label = custom_label, ...)

  if (col_split) {
    lyt <- split_cols_by_multivar(
      lyt = lyt,
      vars = rep(var, length(.stats)),
      varlabels = .labels[.stats],
      extra_args = list(.stats = .stats)
    )
  }
  summarize_row_groups(
    lyt = lyt,
    var = var,
    cfun = a_count_patients_sum_exposure,
    na_str = na_str,
    extra_args = extra_args
  )
}

#' @describeIn summarize_patients_exposure_in_cols Layout-creating function which can take statistics
#'   function arguments and additional format arguments. This function is a wrapper for
#'   [rtables::split_cols_by_multivar()] and [rtables::analyze_colvars()].
#'
#' @param col_split (`flag`)\cr whether the columns should be split. Set to `FALSE` when the required
#'   column split has been done already earlier in the layout pipe.
#'
#' @return
#' * `analyze_patients_exposure_in_cols()` returns a layout object suitable for passing to further
#'   layouting functions, or to [rtables::build_table()]. Adding this function to an `rtable` layout will
#'   add formatted data rows, with the statistics from `s_count_patients_sum_exposure()` arranged in
#'   columns, to the table layout.
#'
#' @note As opposed to [summarize_patients_exposure_in_cols()] which generates content rows,
#'   `analyze_patients_exposure_in_cols()` generates data rows which will _not_ be repeated on multiple
#'   pages when pagination is used.
#'
#' @examples
#' set.seed(1)
#' df <- data.frame(
#'   USUBJID = c(paste("id", seq(1, 12), sep = "")),
#'   ARMCD = c(rep("ARM A", 6), rep("ARM B", 6)),
#'   SEX = c(rep("Female", 6), rep("Male", 6)),
#'   AVAL = as.numeric(sample(seq(1, 20), 12)),
#'   stringsAsFactors = TRUE
#' )
#' adsl <- data.frame(
#'   USUBJID = c(paste("id", seq(1, 12), sep = "")),
#'   ARMCD = c(rep("ARM A", 2), rep("ARM B", 2)),
#'   SEX = c(rep("Female", 2), rep("Male", 2)),
#'   stringsAsFactors = TRUE
#' )
#'
#' lyt <- basic_table() %>%
#'   split_cols_by("ARMCD", split_fun = add_overall_level("Total", first = FALSE)) %>%
#'   summarize_patients_exposure_in_cols(var = "AVAL", col_split = TRUE) %>%
#'   analyze_patients_exposure_in_cols(var = "SEX", col_split = FALSE)
#' result <- build_table(lyt, df = df, alt_counts_df = adsl)
#' result
#'
#' lyt2 <- basic_table() %>%
#'   split_cols_by("ARMCD", split_fun = add_overall_level("Total", first = FALSE)) %>%
#'   summarize_patients_exposure_in_cols(
#'     var = "AVAL", col_split = TRUE,
#'     .stats = "n_patients", custom_label = "some custom label"
#'   ) %>%
#'   analyze_patients_exposure_in_cols(var = "SEX", col_split = FALSE, ex_var = "AVAL")
#' result2 <- build_table(lyt2, df = df, alt_counts_df = adsl)
#' result2
#'
#' lyt3 <- basic_table() %>%
#'   analyze_patients_exposure_in_cols(var = "SEX", col_split = TRUE, ex_var = "AVAL")
#' result3 <- build_table(lyt3, df = df, alt_counts_df = adsl)
#' result3
#'
#' # Adding total levels and custom label
#' lyt4 <- basic_table(
#'   show_colcounts = TRUE
#' ) %>%
#'   analyze_patients_exposure_in_cols(
#'     var = "ARMCD",
#'     col_split = TRUE,
#'     add_total_level = TRUE,
#'     custom_label = "TOTAL"
#'   ) %>%
#'   append_topleft(c("", "Sex"))
#'
#' result4 <- build_table(lyt4, df = df, alt_counts_df = adsl)
#' result4
#'
#' @export
#' @order 2
analyze_patients_exposure_in_cols <- function(lyt, # nolint
                                              var = NULL,
                                              ex_var = "AVAL",
                                              id = "USUBJID",
                                              add_total_level = FALSE,
                                              custom_label = NULL,
                                              col_split = TRUE,
                                              na_str = default_na_str(),
                                              .stats = c("n_patients", "sum_exposure"),
                                              .labels = c(n_patients = "Patients", sum_exposure = "Person time"),
                                              .indent_mods = 0L,
                                              ...) {
  extra_args <- list(
    var = var, ex_var = ex_var, id = id, add_total_level = add_total_level, custom_label = custom_label, ...
  )

  if (col_split) {
    lyt <- split_cols_by_multivar(
      lyt = lyt,
      vars = rep(ex_var, length(.stats)),
      varlabels = .labels[.stats],
      extra_args = list(.stats = .stats)
    )
  }
  lyt <- lyt %>% analyze_colvars(
    afun = a_count_patients_sum_exposure,
    indent_mod = .indent_mods,
    na_str = na_str,
    extra_args = extra_args
  )
  lyt
}

#' Missing data
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Substitute missing data with a string or factor level.
#'
#' @param x (`factor` or `character`)\cr values for which any missing values should be substituted.
#' @param label (`string`)\cr string that missing data should be replaced with.
#'
#' @return `x` with any `NA` values substituted by `label`.
#'
#' @examples
#' explicit_na(c(NA, "a", "b"))
#' is.na(explicit_na(c(NA, "a", "b")))
#'
#' explicit_na(factor(c(NA, "a", "b")))
#' is.na(explicit_na(factor(c(NA, "a", "b"))))
#'
#' explicit_na(sas_na(c("a", "")))
#'
#' @export
explicit_na <- function(x, label = "<Missing>") {
  checkmate::assert_string(label)

  if (is.factor(x)) {
    x <- forcats::fct_na_value_to_level(x, label)
    forcats::fct_drop(x, only = label)
  } else if (is.character(x)) {
    x[is.na(x)] <- label
    x
  } else {
    stop("only factors and character vectors allowed")
  }
}

#' Convert strings to `NA`
#'
#' @description `r lifecycle::badge("stable")`
#'
#' SAS imports missing data as empty strings or strings with whitespaces only. This helper function can be used to
#' convert these values to `NA`s.
#'
#' @inheritParams explicit_na
#' @param empty (`flag`)\cr if `TRUE`, empty strings get replaced by `NA`.
#' @param whitespaces (`flag`)\cr if `TRUE`, strings made from only whitespaces get replaced with `NA`.
#'
#' @return `x` with `""` and/or whitespace-only values substituted by `NA`, depending on the values of
#'   `empty` and `whitespaces`.
#'
#' @examples
#' sas_na(c("1", "", " ", "   ", "b"))
#' sas_na(factor(c("", " ", "b")))
#'
#' is.na(sas_na(c("1", "", " ", "   ", "b")))
#'
#' @export
sas_na <- function(x, empty = TRUE, whitespaces = TRUE) {
  checkmate::assert_flag(empty)
  checkmate::assert_flag(whitespaces)

  if (is.factor(x)) {
    empty_levels <- levels(x) == ""
    if (empty && any(empty_levels)) levels(x)[empty_levels] <- NA

    ws_levels <- grepl("^\\s+$", levels(x))
    if (whitespaces && any(ws_levels)) levels(x)[ws_levels] <- NA

    x
  } else if (is.character(x)) {
    if (empty) x[x == ""] <- NA_character_

    if (whitespaces) x[grepl("^\\s+$", x)] <- NA_character_

    x
  } else {
    stop("only factors and character vectors allowed")
  }
}

#' Odds ratio estimation
#'
#' @description `r lifecycle::badge("stable")`
#'
#' The analyze function [estimate_odds_ratio()] creates a layout element to compare bivariate responses between
#' two groups by estimating an odds ratio and its confidence interval.
#'
#' The primary analysis variable specified by `vars` is the group variable. Additional variables can be included in the
#' analysis via the `variables` argument, which accepts `arm`, an arm variable, and `strata`, a stratification variable.
#' If more than two arm levels are present, they can be combined into two groups using the `groups_list` argument.
#'
#' @inheritParams split_cols_by_groups
#' @inheritParams argument_convention
#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("estimate_odds_ratio")`
#'   to see available statistics for this function.
#'
#' @note
#' * This function uses logistic regression for unstratified analyses, and conditional logistic regression for
#'   stratified analyses. The Wald confidence interval is calculated with the specified confidence level.
#' * For stratified analyses, there is currently no implementation for conditional likelihood confidence intervals,
#'   therefore the likelihood confidence interval is not available as an option.
#' * When `vars` contains only responders or non-responders no odds ratio estimation is possible so the returned
#'   values will be `NA`.
#'
#' @seealso Relevant helper function [h_odds_ratio()].
#'
#' @name odds_ratio
#' @order 1
NULL

#' @describeIn odds_ratio Statistics function which estimates the odds ratio
#'   between a treatment and a control. A `variables` list with `arm` and `strata`
#'   variable names must be passed if a stratified analysis is required.
#'
#' @return
#' * `s_odds_ratio()` returns a named list with the statistics `or_ci`
#'   (containing `est`, `lcl`, and `ucl`) and `n_tot`.
#'
#' @examples
#' # Unstratified analysis.
#' s_odds_ratio(
#'   df = subset(dta, grp == "A"),
#'   .var = "rsp",
#'   .ref_group = subset(dta, grp == "B"),
#'   .in_ref_col = FALSE,
#'   .df_row = dta
#' )
#'
#' # Stratified analysis.
#' s_odds_ratio(
#'   df = subset(dta, grp == "A"),
#'   .var = "rsp",
#'   .ref_group = subset(dta, grp == "B"),
#'   .in_ref_col = FALSE,
#'   .df_row = dta,
#'   variables = list(arm = "grp", strata = "strata")
#' )
#'
#' @export
s_odds_ratio <- function(df,
                         .var,
                         .ref_group,
                         .in_ref_col,
                         .df_row,
                         variables = list(arm = NULL, strata = NULL),
                         conf_level = 0.95,
                         groups_list = NULL) {
  y <- list(or_ci = "", n_tot = "")

  if (!.in_ref_col) {
    assert_proportion_value(conf_level)
    assert_df_with_variables(df, list(rsp = .var))
    assert_df_with_variables(.ref_group, list(rsp = .var))

    if (is.null(variables$strata)) {
      data <- data.frame(
        rsp = c(.ref_group[[.var]], df[[.var]]),
        grp = factor(
          rep(c("ref", "Not-ref"), c(nrow(.ref_group), nrow(df))),
          levels = c("ref", "Not-ref")
        )
      )
      y <- or_glm(data, conf_level = conf_level)
    } else {
      assert_df_with_variables(.df_row, c(list(rsp = .var), variables))

      # The group variable prepared for clogit must be synchronised with combination groups definition.
      if (is.null(groups_list)) {
        ref_grp <- as.character(unique(.ref_group[[variables$arm]]))
        trt_grp <- as.character(unique(df[[variables$arm]]))
        grp <- stats::relevel(factor(.df_row[[variables$arm]]), ref = ref_grp)
      } else {
        # If more than one level in reference col.
        reference <- as.character(unique(.ref_group[[variables$arm]]))
        grp_ref_flag <- vapply(
          X = groups_list,
          FUN.VALUE = TRUE,
          FUN = function(x) all(reference %in% x)
        )
        ref_grp <- names(groups_list)[grp_ref_flag]

        # If more than one level in treatment col.
        treatment <- as.character(unique(df[[variables$arm]]))
        grp_trt_flag <- vapply(
          X = groups_list,
          FUN.VALUE = TRUE,
          FUN = function(x) all(treatment %in% x)
        )
        trt_grp <- names(groups_list)[grp_trt_flag]

        grp <- combine_levels(.df_row[[variables$arm]], levels = reference, new_level = ref_grp)
        grp <- combine_levels(grp, levels = treatment, new_level = trt_grp)
      }

      # The reference level in `grp` must be the same as in the `rtables` column split.
      data <- data.frame(
        rsp = .df_row[[.var]],
        grp = grp,
        strata = interaction(.df_row[variables$strata])
      )
      y_all <- or_clogit(data, conf_level = conf_level)
      checkmate::assert_string(trt_grp)
      checkmate::assert_subset(trt_grp, names(y_all$or_ci))
      y$or_ci <- y_all$or_ci[[trt_grp]]
      y$n_tot <- y_all$n_tot
    }
  }

  y$or_ci <- formatters::with_label(
    x = y$or_ci,
    label = paste0("Odds Ratio (", 100 * conf_level, "% CI)")
  )

  y$n_tot <- formatters::with_label(
    x = y$n_tot,
    label = "Total n"
  )

  y
}

#' @describeIn odds_ratio Formatted analysis function which is used as `afun` in `estimate_odds_ratio()`.
#'
#' @return
#' * `a_odds_ratio()` returns the corresponding list with formatted [rtables::CellValue()].
#'
#' @examples
#' a_odds_ratio(
#'   df = subset(dta, grp == "A"),
#'   .var = "rsp",
#'   .ref_group = subset(dta, grp == "B"),
#'   .in_ref_col = FALSE,
#'   .df_row = dta
#' )
#'
#' @export
a_odds_ratio <- make_afun(
  s_odds_ratio,
  .formats = c(or_ci = "xx.xx (xx.xx - xx.xx)"),
  .indent_mods = c(or_ci = 1L)
)

#' @describeIn odds_ratio Layout-creating function which can take statistics function arguments
#'   and additional format arguments. This function is a wrapper for [rtables::analyze()].
#'
#' @param ... arguments passed to `s_odds_ratio()`.
#'
#' @return
#' * `estimate_odds_ratio()` returns a layout object suitable for passing to further layouting functions,
#'   or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
#'   the statistics from `s_odds_ratio()` to the table layout.
#'
#' @examples
#' set.seed(12)
#' dta <- data.frame(
#'   rsp = sample(c(TRUE, FALSE), 100, TRUE),
#'   grp = factor(rep(c("A", "B"), each = 50), levels = c("A", "B")),
#'   strata = factor(sample(c("C", "D"), 100, TRUE))
#' )
#'
#' l <- basic_table() %>%
#'   split_cols_by(var = "grp", ref_group = "B") %>%
#'   estimate_odds_ratio(vars = "rsp")
#'
#' build_table(l, df = dta)
#'
#' @export
#' @order 2
estimate_odds_ratio <- function(lyt,
                                vars,
                                variables = list(arm = NULL, strata = NULL),
                                conf_level = 0.95,
                                groups_list = NULL,
                                na_str = default_na_str(),
                                nested = TRUE,
                                ...,
                                show_labels = "hidden",
                                table_names = vars,
                                .stats = "or_ci",
                                .formats = NULL,
                                .labels = NULL,
                                .indent_mods = NULL) {
  extra_args <- list(variables = variables, conf_level = conf_level, groups_list = groups_list, ...)

  afun <- make_afun(
    a_odds_ratio,
    .stats = .stats,
    .formats = .formats,
    .labels = .labels,
    .indent_mods = .indent_mods
  )

  analyze(
    lyt,
    vars,
    afun = afun,
    na_str = na_str,
    nested = nested,
    extra_args = extra_args,
    show_labels = show_labels,
    table_names = table_names
  )
}

#' Helper functions for odds ratio estimation
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Functions to calculate odds ratios in [estimate_odds_ratio()].
#'
#' @inheritParams argument_convention
#' @param data (`data.frame`)\cr data frame containing at least the variables `rsp` and `grp`, and optionally
#'   `strata` for [or_clogit()].
#'
#' @return A named `list` of elements `or_ci` and `n_tot`.
#'
#' @seealso [odds_ratio]
#'
#' @name h_odds_ratio
NULL

#' @describeIn h_odds_ratio Estimates the odds ratio based on [stats::glm()]. Note that there must be
#'   exactly 2 groups in `data` as specified by the `grp` variable.
#'
#' @examples
#' # Data with 2 groups.
#' data <- data.frame(
#'   rsp = as.logical(c(1, 1, 0, 1, 0, 0, 1, 1)),
#'   grp = letters[c(1, 1, 1, 2, 2, 2, 1, 2)],
#'   strata = letters[c(1, 2, 1, 2, 2, 2, 1, 2)],
#'   stringsAsFactors = TRUE
#' )
#'
#' # Odds ratio based on glm.
#' or_glm(data, conf_level = 0.95)
#'
#' @export
or_glm <- function(data, conf_level) {
  checkmate::assert_logical(data$rsp)
  assert_proportion_value(conf_level)
  assert_df_with_variables(data, list(rsp = "rsp", grp = "grp"))
  checkmate::assert_multi_class(data$grp, classes = c("factor", "character"))

  data$grp <- as_factor_keep_attributes(data$grp)
  assert_df_with_factors(data, list(val = "grp"), min.levels = 2, max.levels = 2)
  formula <- stats::as.formula("rsp ~ grp")
  model_fit <- stats::glm(
    formula = formula, data = data,
    family = stats::binomial(link = "logit")
  )

  # Note that here we need to discard the intercept.
  or <- exp(stats::coef(model_fit)[-1])
  or_ci <- exp(
    stats::confint.default(model_fit, level = conf_level)[-1, , drop = FALSE]
  )

  values <- stats::setNames(c(or, or_ci), c("est", "lcl", "ucl"))
  n_tot <- stats::setNames(nrow(model_fit$model), "n_tot")

  list(or_ci = values, n_tot = n_tot)
}

#' @describeIn h_odds_ratio Estimates the odds ratio based on [survival::clogit()]. This is done for
#'   the whole data set including all groups, since the results are not the same as when doing
#'   pairwise comparisons between the groups.
#'
#' @examples
#' # Data with 3 groups.
#' data <- data.frame(
#'   rsp = as.logical(c(1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0)),
#'   grp = letters[c(1, 1, 1, 2, 2, 2, 3, 3, 3, 3, 1, 1, 1, 2, 2, 2, 3, 3, 3, 3)],
#'   strata = LETTERS[c(1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2)],
#'   stringsAsFactors = TRUE
#' )
#'
#' # Odds ratio based on stratified estimation by conditional logistic regression.
#' or_clogit(data, conf_level = 0.95)
#'
#' @export
or_clogit <- function(data, conf_level) {
  checkmate::assert_logical(data$rsp)
  assert_proportion_value(conf_level)
  assert_df_with_variables(data, list(rsp = "rsp", grp = "grp", strata = "strata"))
  checkmate::assert_multi_class(data$grp, classes = c("factor", "character"))
  checkmate::assert_multi_class(data$strata, classes = c("factor", "character"))

  data$grp <- as_factor_keep_attributes(data$grp)
  data$strata <- as_factor_keep_attributes(data$strata)

  # Deviation from convention: `survival::strata` must be simply `strata`.
  formula <- stats::as.formula("rsp ~ grp + strata(strata)")
  model_fit <- clogit_with_tryCatch(formula = formula, data = data)

  # Create a list with one set of OR estimates and CI per coefficient, i.e.
  # comparison of one group vs. the reference group.
  coef_est <- stats::coef(model_fit)
  ci_est <- stats::confint(model_fit, level = conf_level)
  or_ci <- list()
  for (coef_name in names(coef_est)) {
    grp_name <- gsub("^grp", "", x = coef_name)
    or_ci[[grp_name]] <- stats::setNames(
      object = exp(c(coef_est[coef_name], ci_est[coef_name, , drop = TRUE])),
      nm = c("est", "lcl", "ucl")
    )
  }
  list(or_ci = or_ci, n_tot = c(n_tot = model_fit$n))
}

#' Helper function to create a new SMQ variable in ADAE by stacking SMQ and/or CQ records.
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Helper function to create a new SMQ variable in ADAE that consists of all adverse events belonging to
#' selected Standardized/Customized queries. The new dataset will only contain records of the adverse events
#' belonging to any of the selected baskets. Remember that `na_str` must match the needed pre-processing
#' done with [df_explicit_na()] to have the desired output.
#'
#' @inheritParams argument_convention
#' @param baskets (`character`)\cr variable names of the selected Standardized/Customized queries.
#' @param smq_varlabel (`string`)\cr a label for the new variable created.
#' @param keys (`character`)\cr names of the key variables to be returned along with the new variable created.
#' @param aag_summary (`data.frame`)\cr containing the SMQ baskets and the levels of interest for the final SMQ
#'   variable. This is useful when there are some levels of interest that are not observed in the `df` dataset.
#'   The two columns of this dataset should be named `basket` and `basket_name`.
#'
#' @return A `data.frame` with variables in `keys` taken from `df` and new variable SMQ containing
#'   records belonging to the baskets selected via the `baskets` argument.
#'
#' @examples
#' adae <- tern_ex_adae[1:20, ] %>% df_explicit_na()
#' h_stack_by_baskets(df = adae)
#'
#' aag <- data.frame(
#'   NAMVAR = c("CQ01NAM", "CQ02NAM", "SMQ01NAM", "SMQ02NAM"),
#'   REFNAME = c(
#'     "D.2.1.5.3/A.1.1.1.1 aesi", "X.9.9.9.9/Y.8.8.8.8 aesi",
#'     "C.1.1.1.3/B.2.2.3.1 aesi", "C.1.1.1.3/B.3.3.3.3 aesi"
#'   ),
#'   SCOPE = c("", "", "BROAD", "BROAD"),
#'   stringsAsFactors = FALSE
#' )
#'
#' basket_name <- character(nrow(aag))
#' cq_pos <- grep("^(CQ).+NAM$", aag$NAMVAR)
#' smq_pos <- grep("^(SMQ).+NAM$", aag$NAMVAR)
#' basket_name[cq_pos] <- aag$REFNAME[cq_pos]
#' basket_name[smq_pos] <- paste0(
#'   aag$REFNAME[smq_pos], "(", aag$SCOPE[smq_pos], ")"
#' )
#'
#' aag_summary <- data.frame(
#'   basket = aag$NAMVAR,
#'   basket_name = basket_name,
#'   stringsAsFactors = TRUE
#' )
#'
#' result <- h_stack_by_baskets(df = adae, aag_summary = aag_summary)
#' all(levels(aag_summary$basket_name) %in% levels(result$SMQ))
#'
#' h_stack_by_baskets(
#'   df = adae,
#'   aag_summary = NULL,
#'   keys = c("STUDYID", "USUBJID", "AEDECOD", "ARM"),
#'   baskets = "SMQ01NAM"
#' )
#'
#' @export
h_stack_by_baskets <- function(df,
                               baskets = grep("^(SMQ|CQ).+NAM$", names(df), value = TRUE),
                               smq_varlabel = "Standardized MedDRA Query",
                               keys = c("STUDYID", "USUBJID", "ASTDTM", "AEDECOD", "AESEQ"),
                               aag_summary = NULL,
                               na_str = "<Missing>") {
  smq_nam <- baskets[startsWith(baskets, "SMQ")]
  # SC corresponding to NAM
  smq_sc <- gsub(pattern = "NAM", replacement = "SC", x = smq_nam, fixed = TRUE)
  smq <- stats::setNames(smq_sc, smq_nam)

  checkmate::assert_character(baskets)
  checkmate::assert_string(smq_varlabel)
  checkmate::assert_data_frame(df)
  checkmate::assert_true(all(startsWith(baskets, "SMQ") | startsWith(baskets, "CQ")))
  checkmate::assert_true(all(endsWith(baskets, "NAM")))
  checkmate::assert_subset(baskets, names(df))
  checkmate::assert_subset(keys, names(df))
  checkmate::assert_subset(smq_sc, names(df))
  checkmate::assert_string(na_str)

  if (!is.null(aag_summary)) {
    assert_df_with_variables(
      df = aag_summary,
      variables = list(val = c("basket", "basket_name"))
    )
    # Warning in case there is no match between `aag_summary$basket` and `baskets` argument.
    # Honestly, I think those should completely match. Target baskets should be the same.
    if (length(intersect(baskets, unique(aag_summary$basket))) == 0) {
      warning("There are 0 baskets in common between aag_summary$basket and `baskets` argument.")
    }
  }

  var_labels <- c(formatters::var_labels(df[, keys]), "SMQ" = smq_varlabel)

  # convert `na_str` records from baskets to NA for the later loop and from wide to long steps
  df[, c(baskets, smq_sc)][df[, c(baskets, smq_sc)] == na_str] <- NA

  if (all(is.na(df[, baskets]))) { # in case there is no level for the target baskets
    df_long <- df[-seq_len(nrow(df)), keys] # we just need an empty data frame keeping all factor levels
  } else {
    # Concatenate SMQxxxNAM with corresponding SMQxxxSC
    df_cnct <- df[, c(keys, baskets[startsWith(baskets, "CQ")])]

    for (nam in names(smq)) {
      sc <- smq[nam] # SMQxxxSC corresponding to SMQxxxNAM
      nam_notna <- !is.na(df[[nam]])
      new_colname <- paste(nam, sc, sep = "_")
      df_cnct[nam_notna, new_colname] <- paste0(df[[nam]], "(", df[[sc]], ")")[nam_notna]
    }

    df_cnct$unique_id <- seq(1, nrow(df_cnct))
    var_cols <- names(df_cnct)[!(names(df_cnct) %in% c(keys, "unique_id"))]
    # have to convert df_cnct from tibble to data frame
    # as it throws a warning otherwise about rownames.
    # tibble do not support rownames and reshape creates rownames

    df_long <- stats::reshape(
      data = as.data.frame(df_cnct),
      varying = var_cols,
      v.names = "SMQ",
      idvar = names(df_cnct)[names(df_cnct) %in% c(keys, "unique_id")],
      direction = "long",
      new.row.names = seq(prod(length(var_cols), nrow(df_cnct)))
    )

    df_long <- df_long[!is.na(df_long[, "SMQ"]), !(names(df_long) %in% c("time", "unique_id"))]
    df_long$SMQ <- as.factor(df_long$SMQ)
  }

  smq_levels <- setdiff(levels(df_long[["SMQ"]]), na_str)

  if (!is.null(aag_summary)) {
    # A warning in case there is no match between df and aag_summary records
    if (length(intersect(smq_levels, unique(aag_summary$basket_name))) == 0) {
      warning("There are 0 basket levels in common between aag_summary$basket_name and df.")
    }
    df_long[["SMQ"]] <- factor(
      df_long[["SMQ"]],
      levels = sort(
        c(
          smq_levels,
          setdiff(unique(aag_summary$basket_name), smq_levels)
        )
      )
    )
  } else {
    all_na_basket_flag <- vapply(df[, baskets], function(x) {
      all(is.na(x))
    }, FUN.VALUE = logical(1))
    all_na_basket <- baskets[all_na_basket_flag]

    df_long[["SMQ"]] <- factor(
      df_long[["SMQ"]],
      levels = sort(c(smq_levels, all_na_basket))
    )
  }
  formatters::var_labels(df_long) <- var_labels
  tibble::tibble(df_long)
}

#' Bland-Altman analysis
#'
#' @description `r lifecycle::badge("experimental")`
#'
#' Functions that use the Bland-Altman method to assess the agreement between two numerical vectors.
#'
#' @inheritParams argument_convention
#' @param y (`numeric`)\cr vector of numbers we want to analyze, to be compared with `x`.
#'
#' @name bland_altman
NULL

#' @describeIn bland_altman Statistics function that compares two numeric vectors using the Bland-Altman method
#'   and calculates a variety of statistics.
#'
#' @return
#' * `s_bland_altman()` returns a named list of the following elements: `df`, `difference_mean`, `ci_mean`,
#'   `difference_sd`, `difference_se`, `upper_agreement_limit`, `lower_agreement_limit`, `agreement_limit_se`,
#'   `upper_agreement_limit_ci`, `lower_agreement_limit_ci`, `t_value`, and `n`.
#'
#' @examples
#' x <- seq(1, 60, 5)
#' y <- seq(5, 50, 4)
#' conf_level <- 0.9
#'
#' # Derive statistics that are needed for Bland-Altman plot
#' s_bland_altman(x, y, conf_level = conf_level)
#'
#' @export
s_bland_altman <- function(x, y, conf_level = 0.95) {
  checkmate::assert_numeric(x, min.len = 1, any.missing = TRUE)
  checkmate::assert_numeric(y, len = length(x), any.missing = TRUE)
  checkmate::assert_numeric(conf_level, lower = 0, upper = 1, any.missing = TRUE)

  alpha <- 1 - conf_level

  ind <- complete.cases(x, y) # use only pairwise complete observations, and check if x and y have the same length
  x <- x[ind]
  y <- y[ind]
  n <- sum(ind) # number of 'observations'

  if (n == 0) {
    stop("there is no valid paired data")
  }

  difference <- x - y # vector of differences
  average <- (x + y) / 2 # vector of means
  difference_mean <- mean(difference) # mean difference
  difference_sd <- sd(difference) # SD of differences
  al <- qnorm(1 - alpha / 2) * difference_sd
  upper_agreement_limit <- difference_mean + al # agreement limits
  lower_agreement_limit <- difference_mean - al

  difference_se <- difference_sd / sqrt(n) # standard error of the mean
  al_se <- difference_sd * sqrt(3) / sqrt(n) # standard error of the agreement limit
  tvalue <- qt(1 - alpha / 2, n - 1) # t value for 95% CI calculation
  difference_mean_ci <- difference_se * tvalue
  al_ci <- al_se * tvalue
  upper_agreement_limit_ci <- c(upper_agreement_limit - al_ci, upper_agreement_limit + al_ci)
  lower_agreement_limit_ci <- c(lower_agreement_limit - al_ci, lower_agreement_limit + al_ci)

  list(
    df = data.frame(average, difference),
    difference_mean = difference_mean,
    ci_mean = difference_mean + c(-1, 1) * difference_mean_ci,
    difference_sd = difference_sd,
    difference_se = difference_se,
    upper_agreement_limit = upper_agreement_limit,
    lower_agreement_limit = lower_agreement_limit,
    agreement_limit_se = al_se,
    upper_agreement_limit_ci = upper_agreement_limit_ci,
    lower_agreement_limit_ci = lower_agreement_limit_ci,
    t_value = tvalue,
    n = n
  )
}

#' @describeIn bland_altman Graphing function that produces a Bland-Altman plot.
#'
#' @return
#' * `g_bland_altman()` returns a `ggplot` Bland-Altman plot.
#'
#' @examples
#' # Create a Bland-Altman plot
#' g_bland_altman(x = x, y = y, conf_level = conf_level)
#'
#' @export
g_bland_altman <- function(x, y, conf_level = 0.95) {
  result_tem <- s_bland_altman(x, y, conf_level = conf_level)
  xpos <- max(result_tem$df$average) * 0.9 + min(result_tem$df$average) * 0.1
  yrange <- diff(range(result_tem$df$difference))

  p <- ggplot(result_tem$df) +
    geom_point(aes(x = average, y = difference), color = "blue") +
    geom_hline(yintercept = result_tem$difference_mean, color = "blue", linetype = 1) +
    geom_hline(yintercept = 0, color = "blue", linetype = 2) +
    geom_hline(yintercept = result_tem$lower_agreement_limit, color = "red", linetype = 2) +
    geom_hline(yintercept = result_tem$upper_agreement_limit, color = "red", linetype = 2) +
    annotate(
      "text",
      x = xpos,
      y = result_tem$lower_agreement_limit + 0.03 * yrange,
      label = "lower limits of agreement",
      color = "red"
    ) +
    annotate(
      "text",
      x = xpos,
      y = result_tem$upper_agreement_limit + 0.03 * yrange,
      label = "upper limits of agreement",
      color = "red"
    ) +
    annotate(
      "text",
      x = xpos,
      y = result_tem$difference_mean + 0.03 * yrange,
      label = "mean of difference between two measures",
      color = "blue"
    ) +
    annotate(
      "text",
      x = xpos,
      y = result_tem$lower_agreement_limit - 0.03 * yrange,
      label = sprintf("%.2f", result_tem$lower_agreement_limit),
      color = "red"
    ) +
    annotate(
      "text",
      x = xpos,
      y = result_tem$upper_agreement_limit - 0.03 * yrange,
      label = sprintf("%.2f", result_tem$upper_agreement_limit),
      color = "red"
    ) +
    annotate(
      "text",
      x = xpos,
      y = result_tem$difference_mean - 0.03 * yrange,
      label = sprintf("%.2f", result_tem$difference_meanm),
      color = "blue"
    ) +
    xlab("Average of two measures") +
    ylab("Difference between two measures")

  return(p)
}

#' Helper functions for Cox proportional hazards regression
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Helper functions used in [fit_coxreg_univar()] and [fit_coxreg_multivar()].
#'
#' @inheritParams argument_convention
#' @inheritParams h_coxreg_univar_extract
#' @inheritParams cox_regression_inter
#' @inheritParams control_coxreg
#'
#' @seealso [cox_regression]
#'
#' @name h_cox_regression
NULL

#' @describeIn h_cox_regression Helper for Cox regression formula. Creates a list of formulas. It is used
#'   internally by [fit_coxreg_univar()] for the comparison of univariate Cox regression models.
#'
#' @return
#' * `h_coxreg_univar_formulas()` returns a `character` vector coercible into formulas (e.g [stats::as.formula()]).
#'
#' @examples
#' # `h_coxreg_univar_formulas`
#'
#' ## Simple formulas.
#' h_coxreg_univar_formulas(
#'   variables = list(
#'     time = "time", event = "status", arm = "armcd", covariates = c("X", "y")
#'   )
#' )
#'
#' ## Addition of an optional strata.
#' h_coxreg_univar_formulas(
#'   variables = list(
#'     time = "time", event = "status", arm = "armcd", covariates = c("X", "y"),
#'     strata = "SITE"
#'   )
#' )
#'
#' ## Inclusion of the interaction term.
#' h_coxreg_univar_formulas(
#'   variables = list(
#'     time = "time", event = "status", arm = "armcd", covariates = c("X", "y"),
#'     strata = "SITE"
#'   ),
#'   interaction = TRUE
#' )
#'
#' ## Only covariates fitted in separate models.
#' h_coxreg_univar_formulas(
#'   variables = list(
#'     time = "time", event = "status", covariates = c("X", "y")
#'   )
#' )
#'
#' @export
h_coxreg_univar_formulas <- function(variables,
                                     interaction = FALSE) {
  checkmate::assert_list(variables, names = "named")
  has_arm <- "arm" %in% names(variables)
  arm_name <- if (has_arm) "arm" else NULL

  checkmate::assert_character(variables$covariates, null.ok = TRUE)

  checkmate::assert_flag(interaction)

  if (!has_arm || is.null(variables$covariates)) {
    checkmate::assert_false(interaction)
  }

  assert_list_of_variables(variables[c(arm_name, "event", "time")])

  if (!is.null(variables$covariates)) {
    forms <- paste0(
      "survival::Surv(", variables$time, ", ", variables$event, ") ~ ",
      ifelse(has_arm, variables$arm, "1"),
      ifelse(interaction, " * ", " + "),
      variables$covariates,
      ifelse(
        !is.null(variables$strata),
        paste0(" + strata(", paste0(variables$strata, collapse = ", "), ")"),
        ""
      )
    )
  } else {
    forms <- NULL
  }
  nams <- variables$covariates
  if (has_arm) {
    ref <- paste0(
      "survival::Surv(", variables$time, ", ", variables$event, ") ~ ",
      variables$arm,
      ifelse(
        !is.null(variables$strata),
        paste0(
          " + strata(", paste0(variables$strata, collapse = ", "), ")"
        ),
        ""
      )
    )
    forms <- c(ref, forms)
    nams <- c("ref", nams)
  }
  stats::setNames(forms, nams)
}

#' @describeIn h_cox_regression Helper for multivariate Cox regression formula. Creates a formulas
#'   string. It is used internally by [fit_coxreg_multivar()] for the comparison of multivariate Cox
#'   regression models. Interactions will not be included in multivariate Cox regression model.
#'
#' @return
#' * `h_coxreg_multivar_formula()` returns a `string` coercible into a formula (e.g [stats::as.formula()]).
#'
#' @examples
#' # `h_coxreg_multivar_formula`
#'
#' h_coxreg_multivar_formula(
#'   variables = list(
#'     time = "AVAL", event = "event", arm = "ARMCD", covariates = c("RACE", "AGE")
#'   )
#' )
#'
#' # Addition of an optional strata.
#' h_coxreg_multivar_formula(
#'   variables = list(
#'     time = "AVAL", event = "event", arm = "ARMCD", covariates = c("RACE", "AGE"),
#'     strata = "SITE"
#'   )
#' )
#'
#' # Example without treatment arm.
#' h_coxreg_multivar_formula(
#'   variables = list(
#'     time = "AVAL", event = "event", covariates = c("RACE", "AGE"),
#'     strata = "SITE"
#'   )
#' )
#'
#' @export
h_coxreg_multivar_formula <- function(variables) {
  checkmate::assert_list(variables, names = "named")
  has_arm <- "arm" %in% names(variables)
  arm_name <- if (has_arm) "arm" else NULL

  checkmate::assert_character(variables$covariates, null.ok = TRUE)

  assert_list_of_variables(variables[c(arm_name, "event", "time")])

  y <- paste0(
    "survival::Surv(", variables$time, ", ", variables$event, ") ~ ",
    ifelse(has_arm, variables$arm, "1")
  )
  if (length(variables$covariates) > 0) {
    y <- paste(y, paste(variables$covariates, collapse = " + "), sep = " + ")
  }
  if (!is.null(variables$strata)) {
    y <- paste0(y, " + strata(", paste0(variables$strata, collapse = ", "), ")")
  }
  y
}

#' @describeIn h_cox_regression Utility function to help tabulate the result of
#'   a univariate Cox regression model.
#'
#' @param effect (`string`)\cr the treatment variable.
#' @param mod (`coxph`)\cr Cox regression model fitted by [survival::coxph()].
#'
#' @return
#' * `h_coxreg_univar_extract()` returns a `data.frame` with variables `effect`, `term`, `term_label`, `level`,
#'   `n`, `hr`, `lcl`, `ucl`, and `pval`.
#'
#' @examples
#' library(survival)
#'
#' dta_simple <- data.frame(
#'   time = c(5, 5, 10, 10, 5, 5, 10, 10),
#'   status = c(0, 0, 1, 0, 0, 1, 1, 1),
#'   armcd = factor(LETTERS[c(1, 1, 1, 1, 2, 2, 2, 2)], levels = c("A", "B")),
#'   var1 = c(45, 55, 65, 75, 55, 65, 85, 75),
#'   var2 = c("F", "M", "F", "M", "F", "M", "F", "U")
#' )
#' mod <- coxph(Surv(time, status) ~ armcd + var1, data = dta_simple)
#' result <- h_coxreg_univar_extract(
#'   effect = "armcd", covar = "armcd", mod = mod, data = dta_simple
#' )
#' result
#'
#' @export
h_coxreg_univar_extract <- function(effect,
                                    covar,
                                    data,
                                    mod,
                                    control = control_coxreg()) {
  checkmate::assert_string(covar)
  checkmate::assert_string(effect)
  checkmate::assert_class(mod, "coxph")
  test_statistic <- c(wald = "Wald", likelihood = "LR")[control$pval_method]

  mod_aov <- muffled_car_anova(mod, test_statistic)
  msum <- summary(mod, conf.int = control$conf_level)
  sum_cox <- broom::tidy(msum)

  # Combine results together.
  effect_aov <- mod_aov[effect, , drop = TRUE]
  pval <- effect_aov[[grep(pattern = "Pr", x = names(effect_aov)), drop = TRUE]]
  sum_main <- sum_cox[grepl(effect, sum_cox$level), ]

  term_label <- if (effect == covar) {
    paste0(
      levels(data[[covar]])[2],
      " vs control (",
      levels(data[[covar]])[1],
      ")"
    )
  } else {
    unname(labels_or_names(data[covar]))
  }
  data.frame(
    effect = ifelse(covar == effect, "Treatment:", "Covariate:"),
    term = covar,
    term_label = term_label,
    level = levels(data[[effect]])[2],
    n = mod[["n"]],
    hr = unname(sum_main["exp(coef)"]),
    lcl = unname(sum_main[grep("lower", names(sum_main))]),
    ucl = unname(sum_main[grep("upper", names(sum_main))]),
    pval = pval,
    stringsAsFactors = FALSE
  )
}

#' @describeIn h_cox_regression Tabulation of multivariate Cox regressions. Utility function to help
#'   tabulate the result of a multivariate Cox regression model for a treatment/covariate variable.
#'
#' @return
#' * `h_coxreg_multivar_extract()` returns a `data.frame` with variables `pval`, `hr`, `lcl`, `ucl`, `level`,
#'   `n`, `term`, and `term_label`.
#'
#' @examples
#' mod <- coxph(Surv(time, status) ~ armcd + var1, data = dta_simple)
#' result <- h_coxreg_multivar_extract(
#'   var = "var1", mod = mod, data = dta_simple
#' )
#' result
#'
#' @export
h_coxreg_multivar_extract <- function(var,
                                      data,
                                      mod,
                                      control = control_coxreg()) {
  test_statistic <- c(wald = "Wald", likelihood = "LR")[control$pval_method]
  mod_aov <- muffled_car_anova(mod, test_statistic)

  msum <- summary(mod, conf.int = control$conf_level)
  sum_anova <- broom::tidy(mod_aov)
  sum_cox <- broom::tidy(msum)

  ret_anova <- sum_anova[sum_anova$term == var, c("term", "p.value")]
  names(ret_anova)[2] <- "pval"
  if (is.factor(data[[var]])) {
    ret_cox <- sum_cox[startsWith(prefix = var, x = sum_cox$level), !(names(sum_cox) %in% "exp(-coef)")]
  } else {
    ret_cox <- sum_cox[(var == sum_cox$level), !(names(sum_cox) %in% "exp(-coef)")]
  }
  names(ret_cox)[1:4] <- c("pval", "hr", "lcl", "ucl")
  varlab <- unname(labels_or_names(data[var]))
  ret_cox$term <- varlab

  if (is.numeric(data[[var]])) {
    ret <- ret_cox
    ret$term_label <- ret$term
  } else if (length(levels(data[[var]])) <= 2) {
    ret_anova$pval <- NA
    ret_anova$term_label <- paste0(varlab, " (reference = ", levels(data[[var]])[1], ")")
    ret_cox$level <- gsub(var, "", ret_cox$level)
    ret_cox$term_label <- ret_cox$level
    ret <- dplyr::bind_rows(ret_anova, ret_cox)
  } else {
    ret_anova$term_label <- paste0(varlab, " (reference = ", levels(data[[var]])[1], ")")
    ret_cox$level <- gsub(var, "", ret_cox$level)
    ret_cox$term_label <- ret_cox$level
    ret <- dplyr::bind_rows(ret_anova, ret_cox)
  }

  as.data.frame(ret)
}

#' Split function to configure risk difference column
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Wrapper function for [rtables::add_combo_levels()] which configures settings for the risk difference
#' column to be added to an `rtables` object. To add a risk difference column to a table, this function
#' should be used as `split_fun` in calls to [rtables::split_cols_by()], followed by setting argument
#' `riskdiff` to `TRUE` in all following analyze function calls.
#'
#' @param arm_x (`string`)\cr name of reference arm to use in risk difference calculations.
#' @param arm_y (`character`)\cr names of one or more arms to compare to reference arm in risk difference
#'   calculations. A new column will be added for each value of `arm_y`.
#' @param col_label (`character`)\cr labels to use when rendering the risk difference column within the table.
#'   If more than one comparison arm is specified in `arm_y`, default labels will specify which two arms are
#'   being compared (reference arm vs. comparison arm).
#' @param pct (`flag`)\cr whether output should be returned as percentages. Defaults to `TRUE`.
#'
#' @return A closure suitable for use as a split function (`split_fun`) within [rtables::split_cols_by()]
#'   when creating a table layout.
#'
#' @seealso [stat_propdiff_ci()] for details on risk difference calculation.
#'
#' @examples
#' adae <- tern_ex_adae
#' adae$AESEV <- factor(adae$AESEV)
#'
#' lyt <- basic_table() %>%
#'   split_cols_by("ARMCD", split_fun = add_riskdiff(arm_x = "ARM A", arm_y = c("ARM B", "ARM C"))) %>%
#'   count_occurrences_by_grade(
#'     var = "AESEV",
#'     riskdiff = TRUE
#'   )
#'
#' tbl <- build_table(lyt, df = adae)
#' tbl
#'
#' @export
add_riskdiff <- function(arm_x,
                         arm_y,
                         col_label = paste0(
                           "Risk Difference (%) (95% CI)", if (length(arm_y) > 1) paste0("\n", arm_x, " vs. ", arm_y)
                         ),
                         pct = TRUE) {
  checkmate::assert_character(arm_x, len = 1)
  checkmate::assert_character(arm_y, min.len = 1)
  checkmate::assert_character(col_label, len = length(arm_y))

  combodf <- tibble::tribble(~valname, ~label, ~levelcombo, ~exargs)
  for (i in seq_len(length(arm_y))) {
    combodf <- rbind(
      combodf,
      tibble::tribble(
        ~valname, ~label, ~levelcombo, ~exargs,
        paste("riskdiff", arm_x, arm_y[i], sep = "_"), col_label[i], c(arm_x, arm_y[i]), list()
      )
    )
  }
  if (pct) combodf$valname <- paste0(combodf$valname, "_pct")
  add_combo_levels(combodf)
}

#' Analysis function to calculate risk difference column values
#'
#' In the risk difference column, this function uses the statistics function associated with `afun` to
#' calculates risk difference values from arm X (reference group) and arm Y. These arms are specified
#' when configuring the risk difference column which is done using the [add_riskdiff()] split function in
#' the previous call to [rtables::split_cols_by()]. For all other columns, applies `afun` as usual. This
#' function utilizes the [stat_propdiff_ci()] function to perform risk difference calculations.
#'
#' @inheritParams argument_convention
#' @param afun (named `list`)\cr a named list containing one name-value pair where the name corresponds to
#'   the name of the statistics function that should be used in calculations and the value is the corresponding
#'   analysis function.
#' @param s_args (named `list`)\cr additional arguments to be passed to the statistics function and analysis
#'   function supplied in `afun`.
#'
#' @return A list of formatted [rtables::CellValue()].
#'
#' @seealso
#' * [stat_propdiff_ci()] for details on risk difference calculation.
#' * Split function [add_riskdiff()] which, when used as `split_fun` within [rtables::split_cols_by()] with
#'   `riskdiff` argument set to `TRUE` in subsequent analyze functions calls, adds a risk difference column
#'   to a table layout.
#'
#' @keywords internal
afun_riskdiff <- function(df,
                          labelstr = "",
                          .var,
                          .N_col, # nolint
                          .N_row, # nolint
                          .df_row,
                          .spl_context,
                          .all_col_counts,
                          .stats,
                          .formats = NULL,
                          .labels = NULL,
                          .indent_mods = NULL,
                          na_str = default_na_str(),
                          afun,
                          s_args = list()) {
  if (!any(grepl("riskdiff", names(.spl_context)))) {
    stop(
      "Please set up levels to use in risk difference calculations using the `add_riskdiff` ",
      "split function within `split_cols_by`. See ?add_riskdiff for details."
    )
  }
  checkmate::assert_list(afun, len = 1, types = "function")
  checkmate::assert_named(afun)
  afun_args <- list(
    .var = .var, .df_row = .df_row, .N_row = .N_row, denom = "N_col", labelstr = labelstr,
    .stats = .stats, .formats = .formats, .labels = .labels, .indent_mods = .indent_mods, na_str = na_str
  )
  afun_args <- afun_args[intersect(names(afun_args), names(as.list(args(afun[[1]]))))]
  if ("denom" %in% names(s_args)) afun_args[["denom"]] <- NULL

  cur_split <- tail(.spl_context$cur_col_split_val[[1]], 1)
  if (!grepl("^riskdiff", cur_split)) {
    # Apply basic afun (no risk difference) in all other columns
    do.call(afun[[1]], args = c(list(df = df, .N_col = .N_col), afun_args, s_args))
  } else {
    arm_x <- strsplit(cur_split, "_")[[1]][2]
    arm_y <- strsplit(cur_split, "_")[[1]][3]
    if (length(.spl_context$cur_col_split[[1]]) > 1) { # Different split name for nested column splits
      arm_spl_x <- gsub("riskdiff", "", paste0(strsplit(.spl_context$cur_col_id[1], "_")[[1]][c(1, 2)], collapse = ""))
      arm_spl_y <- gsub("riskdiff", "", paste0(strsplit(.spl_context$cur_col_id[1], "_")[[1]][c(1, 3)], collapse = ""))
    } else {
      arm_spl_x <- arm_x
      arm_spl_y <- arm_y
    }
    N_col_x <- .all_col_counts[[arm_spl_x]] # nolint
    N_col_y <- .all_col_counts[[arm_spl_y]] # nolint
    cur_var <- tail(.spl_context$cur_col_split[[1]], 1)

    # Apply statistics function to arm X and arm Y data
    s_args <- c(s_args, afun_args[intersect(names(afun_args), names(as.list(args(names(afun)))))])
    s_x <- do.call(names(afun), args = c(list(df = df[df[[cur_var]] == arm_x, ], .N_col = N_col_x), s_args))
    s_y <- do.call(names(afun), args = c(list(df = df[df[[cur_var]] == arm_y, ], .N_col = N_col_y), s_args))

    # Get statistic name and row names
    stat <- ifelse("count_fraction" %in% names(s_x), "count_fraction", "unique")
    if ("flag_variables" %in% names(s_args)) {
      var_nms <- s_args$flag_variables
    } else if (!is.null(names(s_x[[stat]]))) {
      var_nms <- names(s_x[[stat]])
    } else {
      var_nms <- ""
      s_x[[stat]] <- list(s_x[[stat]])
      s_y[[stat]] <- list(s_y[[stat]])
    }

    # Calculate risk difference for each row, repeated if multiple statistics in table
    pct <- tail(strsplit(cur_split, "_")[[1]], 1) == "pct"
    rd_ci <- rep(stat_propdiff_ci(
      lapply(s_x[[stat]], `[`, 1), lapply(s_y[[stat]], `[`, 1),
      N_col_x, N_col_y,
      list_names = var_nms,
      pct = pct
    ), max(1, length(.stats)))

    in_rows(.list = rd_ci, .formats = "xx.x (xx.x - xx.x)", .indent_mods = .indent_mods)
  }
}

#' Control function for risk difference column
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Sets a list of parameters to use when generating a risk (proportion) difference column. Used as input to the
#' `riskdiff` parameter of [tabulate_rsp_subgroups()] and [tabulate_survival_subgroups()].
#'
#' @inheritParams add_riskdiff
#' @param format (`string` or `function`)\cr the format label (string) or formatting function to apply to the risk
#'   difference statistic. See the `3d` string options in [formatters::list_valid_format_labels()] for possible format
#'   strings. Defaults to `"xx.x (xx.x - xx.x)"`.
#'
#' @return A `list` of items with names corresponding to the arguments.
#'
#' @seealso [add_riskdiff()], [tabulate_rsp_subgroups()], and [tabulate_survival_subgroups()].
#'
#' @examples
#' control_riskdiff()
#' control_riskdiff(arm_x = "ARM A", arm_y = "ARM B")
#'
#' @export
control_riskdiff <- function(arm_x = NULL,
                             arm_y = NULL,
                             format = "xx.x (xx.x - xx.x)",
                             col_label = "Risk Difference (%) (95% CI)",
                             pct = TRUE) {
  checkmate::assert_character(arm_x, len = 1, null.ok = TRUE)
  checkmate::assert_character(arm_y, min.len = 1, null.ok = TRUE)
  checkmate::assert_character(format, len = 1)
  checkmate::assert_character(col_label)
  checkmate::assert_flag(pct)

  list(arm_x = arm_x, arm_y = arm_y, format = format, col_label = col_label, pct = pct)
}

#' Encode categorical missing values in a data frame
#'
#' @description `r lifecycle::badge("stable")`
#'
#' This is a helper function to encode missing entries across groups of categorical
#' variables in a data frame.
#'
#' @details Missing entries are those with `NA` or empty strings and will
#'   be replaced with a specified value. If factor variables include missing
#'   values, the missing value will be inserted as the last level.
#'   Similarly, in case character or logical variables should be converted to factors
#'   with the `char_as_factor` or `logical_as_factor` options, the missing values will
#'   be set as the last level.
#'
#' @param data (`data.frame`)\cr data set.
#' @param omit_columns (`character`)\cr names of variables from `data` that should
#'   not be modified by this function.
#' @param char_as_factor (`flag`)\cr whether to convert character variables
#'   in `data` to factors.
#' @param logical_as_factor (`flag`)\cr whether to convert logical variables
#'   in `data` to factors.
#' @param na_level (`string`)\cr string used to replace all `NA` or empty
#'   values inside non-`omit_columns` columns.
#'
#' @return A `data.frame` with the chosen modifications applied.
#'
#' @seealso [sas_na()] and [explicit_na()] for other missing data helper functions.
#'
#' @examples
#' my_data <- data.frame(
#'   u = c(TRUE, FALSE, NA, TRUE),
#'   v = factor(c("A", NA, NA, NA), levels = c("Z", "A")),
#'   w = c("A", "B", NA, "C"),
#'   x = c("D", "E", "F", NA),
#'   y = c("G", "H", "I", ""),
#'   z = c(1, 2, 3, 4),
#'   stringsAsFactors = FALSE
#' )
#'
#' # Example 1
#' # Encode missing values in all character or factor columns.
#' df_explicit_na(my_data)
#' # Also convert logical columns to factor columns.
#' df_explicit_na(my_data, logical_as_factor = TRUE)
#' # Encode missing values in a subset of columns.
#' df_explicit_na(my_data, omit_columns = c("x", "y"))
#'
#' # Example 2
#' # Here we purposefully convert all `M` values to `NA` in the `SEX` variable.
#' # After running `df_explicit_na` the `NA` values are encoded as `<Missing>` but they are not
#' # included when generating `rtables`.
#' adsl <- tern_ex_adsl
#' adsl$SEX[adsl$SEX == "M"] <- NA
#' adsl <- df_explicit_na(adsl)
#'
#' # If you want the `Na` values to be displayed in the table use the `na_level` argument.
#' adsl <- tern_ex_adsl
#' adsl$SEX[adsl$SEX == "M"] <- NA
#' adsl <- df_explicit_na(adsl, na_level = "Missing Values")
#'
#' # Example 3
#' # Numeric variables that have missing values are not altered. This means that any `NA` value in
#' # a numeric variable will not be included in the summary statistics, nor will they be included
#' # in the denominator value for calculating the percent values.
#' adsl <- tern_ex_adsl
#' adsl$AGE[adsl$AGE < 30] <- NA
#' adsl <- df_explicit_na(adsl)
#'
#' @export
df_explicit_na <- function(data,
                           omit_columns = NULL,
                           char_as_factor = TRUE,
                           logical_as_factor = FALSE,
                           na_level = "<Missing>") {
  checkmate::assert_character(omit_columns, null.ok = TRUE, min.len = 1, any.missing = FALSE)
  checkmate::assert_data_frame(data)
  checkmate::assert_flag(char_as_factor)
  checkmate::assert_flag(logical_as_factor)
  checkmate::assert_string(na_level)

  target_vars <- if (is.null(omit_columns)) {
    names(data)
  } else {
    setdiff(names(data), omit_columns) # May have duplicates.
  }
  if (length(target_vars) == 0) {
    return(data)
  }

  l_target_vars <- split(target_vars, target_vars)

  # Makes sure target_vars exist in data and names are not duplicated.
  assert_df_with_variables(data, l_target_vars)

  for (x in target_vars) {
    xi <- data[[x]]
    xi_label <- obj_label(xi)

    # Determine whether to convert character or logical input.
    do_char_conversion <- is.character(xi) && char_as_factor
    do_logical_conversion <- is.logical(xi) && logical_as_factor

    # Pre-convert logical to character to deal correctly with replacing NA
    # values below.
    if (do_logical_conversion) {
      xi <- as.character(xi)
    }

    if (is.factor(xi) || is.character(xi)) {
      # Handle empty strings and NA values.
      xi <- explicit_na(sas_na(xi), label = na_level)

      # Convert to factors if requested for the original type,
      # set na_level as the last value.
      if (do_char_conversion || do_logical_conversion) {
        levels_xi <- setdiff(sort(unique(xi)), na_level)
        if (na_level %in% unique(xi)) {
          levels_xi <- c(levels_xi, na_level)
        }

        xi <- factor(xi, levels = levels_xi)
      }

      data[, x] <- formatters::with_label(xi, label = xi_label)
    }
  }
  return(data)
}

# summarize_glm_count ----------------------------------------------------------
#' Summarize Poisson negative binomial regression
#'
#' @description `r lifecycle::badge("experimental")`
#'
#' Summarize results of a Poisson negative binomial regression.
#' This can be used to analyze count and/or frequency data using a linear model.
#' It is specifically useful for analyzing count data (using the Poisson or Negative
#' Binomial distribution) that is result of a generalized linear model of one (e.g. arm) or more
#' covariates.
#'
#' @inheritParams h_glm_count
#' @inheritParams argument_convention
#' @param rate_mean_method (`character(1)`)\cr method used to estimate the mean odds ratio. Defaults to `emmeans`.
#'   see details for more information.
#' @param scale (`numeric(1)`)\cr linear scaling factor for rate and confidence intervals. Defaults to `1`.
#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("summarize_glm_count")`
#'   to see available statistics for this function.
#'
#' @details
#' `summarize_glm_count()` uses `s_glm_count()` to calculate the statistics for the table. This
#' analysis function uses [h_glm_count()] to estimate the GLM with [stats::glm()] for Poisson and Quasi-Poisson
#' distributions or [MASS::glm.nb()] for Negative Binomial distribution. All methods assume a
#' logarithmic link function.
#'
#' At this point, rates and confidence intervals are estimated from the model using
#' either [emmeans::emmeans()] when `rate_mean_method = "emmeans"` or [h_ppmeans()]
#' when `rate_mean_method = "ppmeans"`.
#'
#' If a reference group is specified while building the table with `split_cols_by(ref_group)`,
#' no rate ratio or `p-value` are calculated. Otherwise, we use [emmeans::contrast()] to
#' calculate the rate ratio and `p-value` for the reference group. Values are always estimated
#' with `method = "trt.vs.ctrl"` and `ref` equal to the first `arm` value.
#'
#' @name summarize_glm_count
NULL

#' @describeIn summarize_glm_count Layout-creating function which can take statistics function arguments
#'   and additional format arguments. This function is a wrapper for [rtables::analyze()].
#'
#' @return
#' * `summarize_glm_count()` returns a layout object suitable for passing to further layouting functions,
#'   or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
#'   the statistics from `s_glm_count()` to the table layout.
#'
#' @examples
#' library(dplyr)
#'
#' anl <- tern_ex_adtte %>% filter(PARAMCD == "TNE")
#' anl$AVAL_f <- as.factor(anl$AVAL)
#'
#' lyt <- basic_table() %>%
#'   split_cols_by("ARM", ref_group = "B: Placebo") %>%
#'   add_colcounts() %>%
#'   analyze_vars(
#'     "AVAL_f",
#'     var_labels = "Number of exacerbations per patient",
#'     .stats = c("count_fraction"),
#'     .formats = c("count_fraction" = "xx (xx.xx%)"),
#'     .labels = c("Number of exacerbations per patient")
#'   ) %>%
#'   summarize_glm_count(
#'     vars = "AVAL",
#'     variables = list(arm = "ARM", offset = "lgTMATRSK", covariates = NULL),
#'     conf_level = 0.95,
#'     distribution = "poisson",
#'     rate_mean_method = "emmeans",
#'     var_labels = "Adjusted (P) exacerbation rate (per year)",
#'     table_names = "adjP",
#'     .stats = c("rate"),
#'     .labels = c(rate = "Rate")
#'   ) %>%
#'   summarize_glm_count(
#'     vars = "AVAL",
#'     variables = list(arm = "ARM", offset = "lgTMATRSK", covariates = c("REGION1")),
#'     conf_level = 0.95,
#'     distribution = "quasipoisson",
#'     rate_mean_method = "ppmeans",
#'     var_labels = "Adjusted (QP) exacerbation rate (per year)",
#'     table_names = "adjQP",
#'     .stats = c("rate", "rate_ci", "rate_ratio", "rate_ratio_ci", "pval"),
#'     .labels = c(
#'       rate = "Rate", rate_ci = "Rate CI", rate_ratio = "Rate Ratio",
#'       rate_ratio_ci = "Rate Ratio CI", pval = "p value"
#'     )
#'   )
#'
#' build_table(lyt = lyt, df = anl)
#'
#' @export
summarize_glm_count <- function(lyt,
                                vars,
                                variables,
                                distribution,
                                conf_level,
                                rate_mean_method = c("emmeans", "ppmeans")[1],
                                weights = stats::weights,
                                scale = 1,
                                var_labels,
                                na_str = default_na_str(),
                                nested = TRUE,
                                ...,
                                show_labels = "visible",
                                table_names = vars,
                                .stats = get_stats("summarize_glm_count"),
                                .formats = NULL,
                                .labels = NULL,
                                .indent_mods = c(
                                  "n" = 0L,
                                  "rate" = 0L,
                                  "rate_ci" = 1L,
                                  "rate_ratio" = 0L,
                                  "rate_ratio_ci" = 1L,
                                  "pval" = 1L
                                )) {
  checkmate::assert_choice(rate_mean_method, c("emmeans", "ppmeans"))

  extra_args <- list(
    variables = variables, distribution = distribution, conf_level = conf_level,
    rate_mean_method = rate_mean_method, weights = weights, scale = scale, ...
  )

  # Selecting parameters following the statistics
  .formats <- get_formats_from_stats(.stats, formats_in = .formats)
  .labels <- get_labels_from_stats(.stats, labels_in = .labels)
  .indent_mods <- get_indents_from_stats(.stats, indents_in = .indent_mods)

  afun <- make_afun(
    s_glm_count,
    .stats = .stats,
    .formats = .formats,
    .labels = .labels,
    .indent_mods = .indent_mods,
    .null_ref_cells = FALSE
  )

  analyze(
    lyt,
    vars,
    var_labels = var_labels,
    show_labels = show_labels,
    table_names = table_names,
    afun = afun,
    na_str = na_str,
    nested = nested,
    extra_args = extra_args
  )
}

#' @describeIn summarize_glm_count Statistics function that produces a named list of results
#'   of the investigated Poisson model.
#'
#' @return
#' * `s_glm_count()` returns a named `list` of 5 statistics:
#'   * `n`: Count of complete sample size for the group.
#'   * `rate`: Estimated event rate per follow-up time.
#'   * `rate_ci`: Confidence level for estimated rate per follow-up time.
#'   * `rate_ratio`: Ratio of event rates in each treatment arm to the reference arm.
#'   * `rate_ratio_ci`: Confidence level for the rate ratio.
#'   * `pval`: p-value.
#'
#' @keywords internal
s_glm_count <- function(df,
                        .var,
                        .df_row,
                        variables,
                        .ref_group,
                        .in_ref_col,
                        distribution,
                        conf_level,
                        rate_mean_method,
                        weights,
                        scale = 1) {
  arm <- variables$arm

  y <- df[[.var]]
  smry_level <- as.character(unique(df[[arm]]))

  # ensure there is only 1 value
  checkmate::assert_scalar(smry_level)

  results <- h_glm_count(
    .var = .var,
    .df_row = .df_row,
    variables = variables,
    distribution = distribution,
    weights
  )

  if (rate_mean_method == "emmeans") {
    emmeans_smry <- summary(results$emmeans_fit, level = conf_level)
  } else if (rate_mean_method == "ppmeans") {
    emmeans_smry <- h_ppmeans(results$glm_fit, .df_row, arm, conf_level)
  }

  emmeans_smry_level <- emmeans_smry[emmeans_smry[[arm]] == smry_level, ]

  # This happens if there is a reference col. No Ratio is calculated?
  if (.in_ref_col) {
    list(
      n = length(y[!is.na(y)]),
      rate = formatters::with_label(
        ifelse(distribution == "negbin", emmeans_smry_level$response * scale, emmeans_smry_level$rate * scale),
        "Adjusted Rate"
      ),
      rate_ci = formatters::with_label(
        c(emmeans_smry_level$asymp.LCL * scale, emmeans_smry_level$asymp.UCL * scale),
        f_conf_level(conf_level)
      ),
      rate_ratio = formatters::with_label(character(), "Adjusted Rate Ratio"),
      rate_ratio_ci = formatters::with_label(character(), f_conf_level(conf_level)),
      pval = formatters::with_label(character(), "p-value")
    )
  } else {
    emmeans_contrasts <- emmeans::contrast(
      results$emmeans_fit,
      method = "trt.vs.ctrl",
      ref = grep(
        as.character(unique(.ref_group[[arm]])),
        as.data.frame(results$emmeans_fit)[[arm]]
      )
    )

    contrasts_smry <- summary(
      emmeans_contrasts,
      infer = TRUE,
      adjust = "none"
    )

    smry_contrasts_level <- contrasts_smry[grepl(smry_level, contrasts_smry$contrast), ]

    list(
      n = length(y[!is.na(y)]),
      rate = formatters::with_label(
        ifelse(distribution == "negbin",
          emmeans_smry_level$response * scale,
          emmeans_smry_level$rate * scale
        ),
        "Adjusted Rate"
      ),
      rate_ci = formatters::with_label(
        c(emmeans_smry_level$asymp.LCL * scale, emmeans_smry_level$asymp.UCL * scale),
        f_conf_level(conf_level)
      ),
      rate_ratio = formatters::with_label(
        smry_contrasts_level$ratio,
        "Adjusted Rate Ratio"
      ),
      rate_ratio_ci = formatters::with_label(
        c(smry_contrasts_level$asymp.LCL, smry_contrasts_level$asymp.UCL),
        f_conf_level(conf_level)
      ),
      pval = formatters::with_label(
        smry_contrasts_level$p.value,
        "p-value"
      )
    )
  }
}
# h_glm_count ------------------------------------------------------------------
#' Helper functions for Poisson models
#'
#' @description `r lifecycle::badge("experimental")`
#'
#' Helper functions that returns the results of [stats::glm()] when Poisson or Quasi-Poisson
#' distributions are needed (see `family` parameter), or [MASS::glm.nb()] for Negative Binomial
#' distributions. Link function for the GLM is `log`.
#'
#' @inheritParams argument_convention
#'
#' @seealso [summarize_glm_count]
#'
#' @name h_glm_count
NULL

#' @describeIn h_glm_count Helper function to return the results of the
#'   selected model (Poisson, Quasi-Poisson, negative binomial).
#'
#' @param .df_row (`data.frame`)\cr dataset that includes all the variables that are called
#'   in `.var` and `variables`.
#' @param variables (named `list` of `string`)\cr list of additional analysis variables, with
#'   expected elements:
#'   * `arm` (`string`)\cr group variable, for which the covariate adjusted means of multiple
#'     groups will be summarized. Specifically, the first level of `arm` variable is taken as the
#'     reference group.
#'   * `covariates` (`character`)\cr a vector that can contain single variable names (such as
#'     `"X1"`), and/or interaction terms indicated by `"X1 * X2"`.
#'   * `offset` (`numeric`)\cr a numeric vector or scalar adding an offset.
#' @param distribution (`character`)\cr a character value specifying the distribution
#'   used in the regression (Poisson, Quasi-Poisson, negative binomial).
#' @param weights (`character`)\cr a character vector specifying weights used
#'   in averaging predictions. Number of weights must equal the number of levels included in the covariates.
#'   Weights option passed to [emmeans::emmeans()].
#'
#' @return
#' * `h_glm_count()` returns the results of the selected model.
#'
#' @keywords internal
h_glm_count <- function(.var,
                        .df_row,
                        variables,
                        distribution,
                        weights) {
  checkmate::assert_subset(distribution, c("poisson", "quasipoisson", "negbin"), empty.ok = FALSE)
  switch(distribution,
    poisson = h_glm_poisson(.var, .df_row, variables, weights),
    quasipoisson = h_glm_quasipoisson(.var, .df_row, variables, weights),
    negbin = h_glm_negbin(.var, .df_row, variables, weights)
  )
}

#' @describeIn h_glm_count Helper function to return results of a Poisson model.
#'
#' @return
#' * `h_glm_poisson()` returns the results of a Poisson model.
#'
#' @keywords internal
h_glm_poisson <- function(.var,
                          .df_row,
                          variables,
                          weights) {
  arm <- variables$arm
  covariates <- variables$covariates
  offset <- .df_row[[variables$offset]]

  formula <- stats::as.formula(paste0(
    .var, " ~ ",
    " + ",
    paste(covariates, collapse = " + "),
    " + ",
    arm
  ))

  glm_fit <- stats::glm(
    formula = formula,
    offset = offset,
    data = .df_row,
    family = stats::poisson(link = "log")
  )

  emmeans_fit <- emmeans::emmeans(
    glm_fit,
    specs = arm,
    data = .df_row,
    type = "response",
    offset = 0,
    weights = weights
  )

  list(
    glm_fit = glm_fit,
    emmeans_fit = emmeans_fit
  )
}

#' @describeIn h_glm_count Helper function to return results of a Quasi-Poisson model.
#'
#' @return
#' * `h_glm_quasipoisson()` returns the results of a Quasi-Poisson model.
#'
#' @keywords internal
h_glm_quasipoisson <- function(.var,
                               .df_row,
                               variables,
                               weights) {
  arm <- variables$arm
  covariates <- variables$covariates
  offset <- .df_row[[variables$offset]]

  formula <- stats::as.formula(paste0(
    .var, " ~ ",
    " + ",
    paste(covariates, collapse = " + "),
    " + ",
    arm
  ))

  glm_fit <- stats::glm(
    formula = formula,
    offset = offset,
    data = .df_row,
    family = stats::quasipoisson(link = "log")
  )

  emmeans_fit <- emmeans::emmeans(
    glm_fit,
    specs = arm,
    data = .df_row,
    type = "response",
    offset = 0,
    weights = weights
  )

  list(
    glm_fit = glm_fit,
    emmeans_fit = emmeans_fit
  )
}

#' @describeIn h_glm_count Helper function to return results of a negative binomial model.
#'
#' @return
#' * `h_glm_negbin()` returns the results of a negative binomial model.
#'
#' @keywords internal
h_glm_negbin <- function(.var,
                         .df_row,
                         variables,
                         weights) {
  arm <- variables$arm
  covariates <- variables$covariates

  formula <- stats::as.formula(paste0(
    .var, " ~ ",
    " + ",
    paste(covariates, collapse = " + "),
    " + ",
    arm
  ))

  glm_fit <- MASS::glm.nb(
    formula = formula,
    data = .df_row,
    link = "log"
  )

  emmeans_fit <- emmeans::emmeans(
    glm_fit,
    specs = arm,
    data = .df_row,
    type = "response",
    offset = 0,
    weights = weights
  )

  list(
    glm_fit = glm_fit,
    emmeans_fit = emmeans_fit
  )
}

# h_ppmeans --------------------------------------------------------------------
#' Function to return the estimated means using predicted probabilities
#'
#' @description
#' For each arm level, the predicted mean rate is calculated using the fitted model object, with `newdata`
#' set to the result of `stats::model.frame`, a reconstructed data or the original data, depending on the
#' object formula (coming from the fit). The confidence interval is derived using the `conf_level` parameter.
#'
#' @param obj (`glm.fit`)\cr fitted model object used to derive the mean rate estimates in each treatment arm.
#' @param .df_row (`data.frame`)\cr dataset that includes all the variables that are called in `.var` and `variables`.
#' @param arm (`string`)\cr group variable, for which the covariate adjusted means of multiple groups will be
#'   summarized. Specifically, the first level of `arm` variable is taken as the reference group.
#' @param conf_level (`proportion`)\cr value used to derive the confidence interval for the rate.
#'
#' @return
#' * `h_ppmeans()` returns the estimated means.
#'
#' @seealso [summarize_glm_count()].
#'
#' @export
h_ppmeans <- function(obj, .df_row, arm, conf_level) {
  alpha <- 1 - conf_level
  p <- 1 - alpha / 2

  arm_levels <- levels(.df_row[[arm]])

  out <- lapply(arm_levels, function(lev) {
    temp <- .df_row
    temp[[arm]] <- factor(lev, levels = arm_levels)

    mf <- stats::model.frame(obj$formula, data = temp)
    X <- stats::model.matrix(obj$formula, data = mf) # nolint

    rate <- stats::predict(obj, newdata = mf, type = "response")
    rate_hat <- mean(rate)

    zz <- colMeans(rate * X)
    se <- sqrt(as.numeric(t(zz) %*% stats::vcov(obj) %*% zz))
    rate_lwr <- rate_hat * exp(-stats::qnorm(p) * se / rate_hat)
    rate_upr <- rate_hat * exp(stats::qnorm(p) * se / rate_hat)

    c(rate_hat, rate_lwr, rate_upr)
  })

  names(out) <- arm_levels
  out <- do.call(rbind, out)
  if ("negbin" %in% class(obj)) {
    colnames(out) <- c("response", "asymp.LCL", "asymp.UCL")
  } else {
    colnames(out) <- c("rate", "asymp.LCL", "asymp.UCL")
  }
  out <- as.data.frame(out)
  out[[arm]] <- rownames(out)
  out
}

#' Analyze numeric variables in columns
#'
#' @description `r lifecycle::badge("experimental")`
#'
#' The layout-creating function [analyze_vars_in_cols()] creates a layout element to generate a column-wise
#' analysis table.
#'
#' This function sets the analysis methods as column labels and is a wrapper for [rtables::analyze_colvars()].
#' It was designed principally for PK tables.
#'
#' @inheritParams argument_convention
#' @inheritParams rtables::analyze_colvars
#' @param imp_rule (`string` or `NULL`)\cr imputation rule setting. Defaults to `NULL` for no imputation rule. Can
#'   also be `"1/3"` to implement 1/3 imputation rule or `"1/2"` to implement 1/2 imputation rule. In order
#'   to use an imputation rule, the `avalcat_var` argument must be specified. See [imputation_rule()]
#'   for more details on imputation.
#' @param avalcat_var (`string`)\cr if `imp_rule` is not `NULL`, name of variable that indicates whether a
#'   row in the data corresponds to an analysis value in category `"BLQ"`, `"LTR"`, `"<PCLLOQ"`, or none of
#'   the above (defaults to `"AVALCAT1"`). Variable must be present in the data and should match the variable
#'   used to calculate the `n_blq` statistic (if included in `.stats`).
#' @param cache (`flag`)\cr whether to store computed values in a temporary caching environment. This will
#'   speed up calculations in large tables, but should be set to `FALSE` if the same `rtable` layout is
#'   used for multiple tables with different data. Defaults to `FALSE`.
#' @param row_labels (`character`)\cr as this function works in columns space, usually `.labels`
#'   character vector applies on the column space. You can change the row labels by defining this
#'   parameter to a named character vector with names corresponding to the split values. It defaults
#'   to `NULL` and if it contains only one `string`, it will duplicate that as a row label.
#' @param do_summarize_row_groups (`flag`)\cr defaults to `FALSE` and applies the analysis to the current
#'   label rows. This is a wrapper of [rtables::summarize_row_groups()] and it can accept `labelstr`
#'   to define row labels. This behavior is not supported as we never need to overload row labels.
#' @param split_col_vars (`flag`)\cr defaults to `TRUE` and puts the analysis results onto the columns.
#'   This option allows you to add multiple instances of this functions, also in a nested fashion,
#'   without adding more splits. This split must happen only one time on a single layout.
#'
#' @return
#' A layout object suitable for passing to further layouting functions, or to [rtables::build_table()].
#' Adding this function to an `rtable` layout will summarize the given variables, arrange the output
#' in columns, and add it to the table layout.
#'
#' @note
#' * This is an experimental implementation of [rtables::summarize_row_groups()] and [rtables::analyze_colvars()]
#'   that may be subjected to changes as `rtables` extends its support to more complex analysis pipelines in the
#'   column space. We encourage users to read the examples carefully and file issues for different use cases.
#' * In this function, `labelstr` behaves atypically. If `labelstr = NULL` (the default), row labels are assigned
#'   automatically as the split values if `do_summarize_row_groups = FALSE` (the default), and as the group label
#'   if `do_summarize_row_groups = TRUE`.
#'
#' @seealso [analyze_vars()], [rtables::analyze_colvars()].
#'
#' @examples
#' library(dplyr)
#'
#' # Data preparation
#' adpp <- tern_ex_adpp %>% h_pkparam_sort()
#'
#' lyt <- basic_table() %>%
#'   split_rows_by(var = "STRATA1", label_pos = "topleft") %>%
#'   split_rows_by(
#'     var = "SEX",
#'     label_pos = "topleft",
#'     child_labels = "hidden"
#'   ) %>% # Removes duplicated labels
#'   analyze_vars_in_cols(vars = "AGE")
#' result <- build_table(lyt = lyt, df = adpp)
#' result
#'
#' # By selecting just some statistics and ad-hoc labels
#' lyt <- basic_table() %>%
#'   split_rows_by(var = "ARM", label_pos = "topleft") %>%
#'   split_rows_by(
#'     var = "SEX",
#'     label_pos = "topleft",
#'     child_labels = "hidden",
#'     split_fun = drop_split_levels
#'   ) %>%
#'   analyze_vars_in_cols(
#'     vars = "AGE",
#'     .stats = c("n", "cv", "geom_mean"),
#'     .labels = c(
#'       n = "aN",
#'       cv = "aCV",
#'       geom_mean = "aGeomMean"
#'     )
#'   )
#' result <- build_table(lyt = lyt, df = adpp)
#' result
#'
#' # Changing row labels
#' lyt <- basic_table() %>%
#'   analyze_vars_in_cols(
#'     vars = "AGE",
#'     row_labels = "some custom label"
#'   )
#' result <- build_table(lyt, df = adpp)
#' result
#'
#' # Pharmacokinetic parameters
#' lyt <- basic_table() %>%
#'   split_rows_by(
#'     var = "TLG_DISPLAY",
#'     split_label = "PK Parameter",
#'     label_pos = "topleft",
#'     child_labels = "hidden"
#'   ) %>%
#'   analyze_vars_in_cols(
#'     vars = "AVAL"
#'   )
#' result <- build_table(lyt, df = adpp)
#' result
#'
#' # Multiple calls (summarize label and analyze underneath)
#' lyt <- basic_table() %>%
#'   split_rows_by(
#'     var = "TLG_DISPLAY",
#'     split_label = "PK Parameter",
#'     label_pos = "topleft"
#'   ) %>%
#'   analyze_vars_in_cols(
#'     vars = "AVAL",
#'     do_summarize_row_groups = TRUE # does a summarize level
#'   ) %>%
#'   split_rows_by("SEX",
#'     child_labels = "hidden",
#'     label_pos = "topleft"
#'   ) %>%
#'   analyze_vars_in_cols(
#'     vars = "AVAL",
#'     split_col_vars = FALSE # avoids re-splitting the columns
#'   )
#' result <- build_table(lyt, df = adpp)
#' result
#'
#' @export
analyze_vars_in_cols <- function(lyt,
                                 vars,
                                 ...,
                                 .stats = c(
                                   "n",
                                   "mean",
                                   "sd",
                                   "se",
                                   "cv",
                                   "geom_cv"
                                 ),
                                 .labels = c(
                                   n = "n",
                                   mean = "Mean",
                                   sd = "SD",
                                   se = "SE",
                                   cv = "CV (%)",
                                   geom_cv = "CV % Geometric Mean"
                                 ),
                                 row_labels = NULL,
                                 do_summarize_row_groups = FALSE,
                                 split_col_vars = TRUE,
                                 imp_rule = NULL,
                                 avalcat_var = "AVALCAT1",
                                 cache = FALSE,
                                 .indent_mods = NULL,
                                 na_str = default_na_str(),
                                 nested = TRUE,
                                 .formats = NULL,
                                 .aligns = NULL) {
  extra_args <- list(...)

  checkmate::assert_string(na_str, na.ok = TRUE, null.ok = TRUE)
  checkmate::assert_character(row_labels, null.ok = TRUE)
  checkmate::assert_int(.indent_mods, null.ok = TRUE)
  checkmate::assert_flag(nested)
  checkmate::assert_flag(split_col_vars)
  checkmate::assert_flag(do_summarize_row_groups)

  # Filtering
  met_grps <- paste0("analyze_vars", c("_numeric", "_counts"))
  .stats <- get_stats(met_grps, stats_in = .stats)
  formats_v <- get_formats_from_stats(stats = .stats, formats_in = .formats)
  labels_v <- get_labels_from_stats(stats = .stats, labels_in = .labels)
  if ("control" %in% names(extra_args)) labels_v <- labels_v %>% labels_use_control(extra_args[["control"]], .labels)

  # Check for vars in the case that one or more are used
  if (length(vars) == 1) {
    vars <- rep(vars, length(.stats))
  } else if (length(vars) != length(.stats)) {
    stop(
      "Analyzed variables (vars) does not have the same ",
      "number of elements of specified statistics (.stats)."
    )
  }

  if (split_col_vars) {
    # Checking there is not a previous identical column split
    clyt <- tail(clayout(lyt), 1)[[1]]

    dummy_lyt <- split_cols_by_multivar(
      lyt = basic_table(),
      vars = vars,
      varlabels = labels_v
    )

    if (any(sapply(clyt, identical, y = get_last_col_split(dummy_lyt)))) {
      stop(
        "Column split called again with the same values. ",
        "This can create many unwanted columns. Please consider adding ",
        "split_col_vars = FALSE to the last call of ",
        deparse(sys.calls()[[sys.nframe() - 1]]), "."
      )
    }

    # Main col split
    lyt <- split_cols_by_multivar(
      lyt = lyt,
      vars = vars,
      varlabels = labels_v
    )
  }

  env <- new.env() # create caching environment

  if (do_summarize_row_groups) {
    if (length(unique(vars)) > 1) {
      stop("When using do_summarize_row_groups only one label level var should be inserted.")
    }

    # Function list for do_summarize_row_groups. Slightly different handling of labels
    cfun_list <- Map(
      function(stat, use_cache, cache_env) {
        function(u, .spl_context, labelstr, .df_row, ...) {
          # Statistic
          var_row_val <- paste(
            gsub("\\._\\[\\[[0-9]+\\]\\]_\\.", "", paste(tail(.spl_context$cur_col_split_val, 1)[[1]], collapse = "_")),
            paste(.spl_context$value, collapse = "_"),
            sep = "_"
          )
          if (use_cache) {
            if (is.null(cache_env[[var_row_val]])) cache_env[[var_row_val]] <- s_summary(u, ...)
            x_stats <- cache_env[[var_row_val]]
          } else {
            x_stats <- s_summary(u, ...)
          }

          if (is.null(imp_rule) || !stat %in% c("mean", "sd", "cv", "geom_mean", "geom_cv", "median", "min", "max")) {
            res <- x_stats[[stat]]
          } else {
            timept <- as.numeric(gsub(".*?([0-9\\.]+).*", "\\1", tail(.spl_context$value, 1)))
            res_imp <- imputation_rule(
              .df_row, x_stats, stat,
              imp_rule = imp_rule,
              post = grepl("Predose", tail(.spl_context$value, 1)) || timept > 0,
              avalcat_var = avalcat_var
            )
            res <- res_imp[["val"]]
            na_str <- res_imp[["na_str"]]
          }

          # Label check and replacement
          if (length(row_labels) > 1) {
            if (!(labelstr %in% names(row_labels))) {
              stop(
                "Replacing the labels in do_summarize_row_groups needs a named vector",
                "that contains the split values. In the current split variable ",
                .spl_context$split[nrow(.spl_context)],
                " the labelstr value (split value by default) ", labelstr, " is not in",
                " row_labels names: ", names(row_labels)
              )
            }
            lbl <- unlist(row_labels[labelstr])
          } else {
            lbl <- labelstr
          }

          # Cell creation
          rcell(res,
            label = lbl,
            format = formats_v[names(formats_v) == stat][[1]],
            format_na_str = na_str,
            indent_mod = ifelse(is.null(.indent_mods), 0L, .indent_mods),
            align = .aligns
          )
        }
      },
      stat = .stats,
      use_cache = cache,
      cache_env = replicate(length(.stats), env)
    )

    # Main call to rtables
    summarize_row_groups(
      lyt = lyt,
      var = unique(vars),
      cfun = cfun_list,
      na_str = na_str,
      extra_args = extra_args
    )
  } else {
    # Function list for analyze_colvars
    afun_list <- Map(
      function(stat, use_cache, cache_env) {
        function(u, .spl_context, .df_row, ...) {
          # Main statistics
          var_row_val <- paste(
            gsub("\\._\\[\\[[0-9]+\\]\\]_\\.", "", paste(tail(.spl_context$cur_col_split_val, 1)[[1]], collapse = "_")),
            paste(.spl_context$value, collapse = "_"),
            sep = "_"
          )
          if (use_cache) {
            if (is.null(cache_env[[var_row_val]])) cache_env[[var_row_val]] <- s_summary(u, ...)
            x_stats <- cache_env[[var_row_val]]
          } else {
            x_stats <- s_summary(u, ...)
          }

          if (is.null(imp_rule) || !stat %in% c("mean", "sd", "cv", "geom_mean", "geom_cv", "median", "min", "max")) {
            res <- x_stats[[stat]]
          } else {
            timept <- as.numeric(gsub(".*?([0-9\\.]+).*", "\\1", tail(.spl_context$value, 1)))
            res_imp <- imputation_rule(
              .df_row, x_stats, stat,
              imp_rule = imp_rule,
              post = grepl("Predose", tail(.spl_context$value, 1)) || timept > 0,
              avalcat_var = avalcat_var
            )
            res <- res_imp[["val"]]
            na_str <- res_imp[["na_str"]]
          }

          if (is.list(res)) {
            if (length(res) > 1) {
              stop("The analyzed column produced more than one category of results.")
            } else {
              res <- unlist(res)
            }
          }

          # Label from context
          label_from_context <- .spl_context$value[nrow(.spl_context)]

          # Label switcher
          if (is.null(row_labels)) {
            lbl <- label_from_context
          } else {
            if (length(row_labels) > 1) {
              if (!(label_from_context %in% names(row_labels))) {
                stop(
                  "Replacing the labels in do_summarize_row_groups needs a named vector",
                  "that contains the split values. In the current split variable ",
                  .spl_context$split[nrow(.spl_context)],
                  " the split value ", label_from_context, " is not in",
                  " row_labels names: ", names(row_labels)
                )
              }
              lbl <- unlist(row_labels[label_from_context])
            } else {
              lbl <- row_labels
            }
          }

          # Cell creation
          rcell(res,
            label = lbl,
            format = formats_v[names(formats_v) == stat][[1]],
            format_na_str = na_str,
            indent_mod = ifelse(is.null(.indent_mods), 0L, .indent_mods),
            align = .aligns
          )
        }
      },
      stat = .stats,
      use_cache = cache,
      cache_env = replicate(length(.stats), env)
    )

    # Main call to rtables
    analyze_colvars(lyt,
      afun = afun_list,
      na_str = na_str,
      nested = nested,
      extra_args = extra_args
    )
  }
}

# Helper function
get_last_col_split <- function(lyt) {
  tail(tail(clayout(lyt), 1)[[1]], 1)[[1]]
}

#' Convert list of groups to a data frame
#'
#' This converts a list of group levels into a data frame format which is expected by [rtables::add_combo_levels()].
#'
#' @param groups_list (named `list` of `character`)\cr specifies the new group levels via the names and the
#'   levels that belong to it in the character vectors that are elements of the list.
#'
#' @return A `tibble` in the required format.
#'
#' @examples
#' grade_groups <- list(
#'   "Any Grade (%)" = c("1", "2", "3", "4", "5"),
#'   "Grade 3-4 (%)" = c("3", "4"),
#'   "Grade 5 (%)" = "5"
#' )
#' groups_list_to_df(grade_groups)
#'
#' @export
groups_list_to_df <- function(groups_list) {
  checkmate::assert_list(groups_list, names = "named")
  lapply(groups_list, checkmate::assert_character)
  tibble::tibble(
    valname = make_names(names(groups_list)),
    label = names(groups_list),
    levelcombo = unname(groups_list),
    exargs = replicate(length(groups_list), list())
  )
}

#' Reference and treatment group combination
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Facilitate the re-combination of groups divided as reference and treatment groups; it helps in arranging groups of
#' columns in the `rtables` framework and teal modules.
#'
#' @param fct (`factor`)\cr the variable with levels which needs to be grouped.
#' @param ref (`character`)\cr the reference level(s).
#' @param collapse (`string`)\cr a character string to separate `fct` and `ref`.
#'
#' @return A `list` with first item `ref` (reference) and second item `trt` (treatment).
#'
#' @examples
#' groups <- combine_groups(
#'   fct = DM$ARM,
#'   ref = c("B: Placebo")
#' )
#'
#' basic_table() %>%
#'   split_cols_by_groups("ARM", groups) %>%
#'   add_colcounts() %>%
#'   analyze_vars("AGE") %>%
#'   build_table(DM)
#'
#' @export
combine_groups <- function(fct,
                           ref = NULL,
                           collapse = "/") {
  checkmate::assert_string(collapse)
  checkmate::assert_character(ref, min.chars = 1, any.missing = FALSE, null.ok = TRUE)
  checkmate::assert_multi_class(fct, classes = c("factor", "character"))

  fct <- as_factor_keep_attributes(fct)

  group_levels <- levels(fct)
  if (is.null(ref)) {
    ref <- group_levels[1]
  } else {
    checkmate::assert_subset(ref, group_levels)
  }

  groups <- list(
    ref = group_levels[group_levels %in% ref],
    trt = group_levels[!group_levels %in% ref]
  )
  stats::setNames(groups, nm = lapply(groups, paste, collapse = collapse))
}

#' Split columns by groups of levels
#'
#' @description `r lifecycle::badge("stable")`
#'
#' @inheritParams argument_convention
#' @inheritParams groups_list_to_df
#' @param ... additional arguments to [rtables::split_cols_by()] in order. For instance, to
#'   control formats (`format`), add a joint column for all groups (`incl_all`).
#'
#' @return A layout object suitable for passing to further layouting functions. Adding
#'   this function to an `rtable` layout will add a column split including the given
#'   groups to the table layout.
#'
#' @seealso [rtables::split_cols_by()]
#'
#' @examples
#' # 1 - Basic use
#'
#' # Without group combination `split_cols_by_groups` is
#' # equivalent to [rtables::split_cols_by()].
#' basic_table() %>%
#'   split_cols_by_groups("ARM") %>%
#'   add_colcounts() %>%
#'   analyze("AGE") %>%
#'   build_table(DM)
#'
#' # Add a reference column.
#' basic_table() %>%
#'   split_cols_by_groups("ARM", ref_group = "B: Placebo") %>%
#'   add_colcounts() %>%
#'   analyze(
#'     "AGE",
#'     afun = function(x, .ref_group, .in_ref_col) {
#'       if (.in_ref_col) {
#'         in_rows("Diff Mean" = rcell(NULL))
#'       } else {
#'         in_rows("Diff Mean" = rcell(mean(x) - mean(.ref_group), format = "xx.xx"))
#'       }
#'     }
#'   ) %>%
#'   build_table(DM)
#'
#' # 2 - Adding group specification
#'
#' # Manual preparation of the groups.
#' groups <- list(
#'   "Arms A+B" = c("A: Drug X", "B: Placebo"),
#'   "Arms A+C" = c("A: Drug X", "C: Combination")
#' )
#'
#' # Use of split_cols_by_groups without reference column.
#' basic_table() %>%
#'   split_cols_by_groups("ARM", groups) %>%
#'   add_colcounts() %>%
#'   analyze("AGE") %>%
#'   build_table(DM)
#'
#' # Including differentiated output in the reference column.
#' basic_table() %>%
#'   split_cols_by_groups("ARM", groups_list = groups, ref_group = "Arms A+B") %>%
#'   analyze(
#'     "AGE",
#'     afun = function(x, .ref_group, .in_ref_col) {
#'       if (.in_ref_col) {
#'         in_rows("Diff. of Averages" = rcell(NULL))
#'       } else {
#'         in_rows("Diff. of Averages" = rcell(mean(x) - mean(.ref_group), format = "xx.xx"))
#'       }
#'     }
#'   ) %>%
#'   build_table(DM)
#'
#' # 3 - Binary list dividing factor levels into reference and treatment
#'
#' # `combine_groups` defines reference and treatment.
#' groups <- combine_groups(
#'   fct = DM$ARM,
#'   ref = c("A: Drug X", "B: Placebo")
#' )
#' groups
#'
#' # Use group definition without reference column.
#' basic_table() %>%
#'   split_cols_by_groups("ARM", groups_list = groups) %>%
#'   add_colcounts() %>%
#'   analyze("AGE") %>%
#'   build_table(DM)
#'
#' # Use group definition with reference column (first item of groups).
#' basic_table() %>%
#'   split_cols_by_groups("ARM", groups, ref_group = names(groups)[1]) %>%
#'   add_colcounts() %>%
#'   analyze(
#'     "AGE",
#'     afun = function(x, .ref_group, .in_ref_col) {
#'       if (.in_ref_col) {
#'         in_rows("Diff Mean" = rcell(NULL))
#'       } else {
#'         in_rows("Diff Mean" = rcell(mean(x) - mean(.ref_group), format = "xx.xx"))
#'       }
#'     }
#'   ) %>%
#'   build_table(DM)
#'
#' @export
split_cols_by_groups <- function(lyt,
                                 var,
                                 groups_list = NULL,
                                 ref_group = NULL,
                                 ...) {
  if (is.null(groups_list)) {
    split_cols_by(
      lyt = lyt,
      var = var,
      ref_group = ref_group,
      ...
    )
  } else {
    groups_df <- groups_list_to_df(groups_list)
    if (!is.null(ref_group)) {
      ref_group <- groups_df$valname[groups_df$label == ref_group]
    }
    split_cols_by(
      lyt = lyt,
      var = var,
      split_fun = add_combo_levels(groups_df, keep_levels = groups_df$valname),
      ref_group = ref_group,
      ...
    )
  }
}

#' Combine counts
#'
#' Simplifies the estimation of column counts, especially when group combination is required.
#'
#' @inheritParams combine_groups
#' @inheritParams groups_list_to_df
#'
#' @return A `vector` of column counts.
#'
#' @seealso [combine_groups()]
#'
#' @examples
#' ref <- c("A: Drug X", "B: Placebo")
#' groups <- combine_groups(fct = DM$ARM, ref = ref)
#'
#' col_counts <- combine_counts(
#'   fct = DM$ARM,
#'   groups_list = groups
#' )
#'
#' basic_table() %>%
#'   split_cols_by_groups("ARM", groups) %>%
#'   add_colcounts() %>%
#'   analyze_vars("AGE") %>%
#'   build_table(DM, col_counts = col_counts)
#'
#' ref <- "A: Drug X"
#' groups <- combine_groups(fct = DM$ARM, ref = ref)
#' col_counts <- combine_counts(
#'   fct = DM$ARM,
#'   groups_list = groups
#' )
#'
#' basic_table() %>%
#'   split_cols_by_groups("ARM", groups) %>%
#'   add_colcounts() %>%
#'   analyze_vars("AGE") %>%
#'   build_table(DM, col_counts = col_counts)
#'
#' @export
combine_counts <- function(fct, groups_list = NULL) {
  checkmate::assert_multi_class(fct, classes = c("factor", "character"))

  fct <- as_factor_keep_attributes(fct)

  if (is.null(groups_list)) {
    y <- table(fct)
    y <- stats::setNames(as.numeric(y), nm = dimnames(y)[[1]])
  } else {
    y <- vapply(
      X = groups_list,
      FUN = function(x) sum(table(fct)[x]),
      FUN.VALUE = 1
    )
  }
  y
}

#' Subgroup treatment effect pattern (STEP) fit for survival outcome
#'
#' @description `r lifecycle::badge("stable")`
#'
#' This fits the subgroup treatment effect pattern (STEP) models for a survival outcome. The treatment arm
#' variable must have exactly 2 levels, where the first one is taken as reference and the estimated
#' hazard ratios are for the comparison of the second level vs. the first one.
#'
#' The model which is fit is:
#'
#' `Surv(time, event) ~ arm * poly(biomarker, degree) + covariates + strata(strata)`
#'
#' where `degree` is specified by `control_step()`.
#'
#' @inheritParams argument_convention
#' @param variables (named `list` of `character`)\cr list of analysis variables: needs `time`, `event`,
#'   `arm`, `biomarker`, and optional `covariates` and `strata`.
#' @param control (named `list`)\cr combined control list from [control_step()] and [control_coxph()].
#'
#' @return A matrix of class `step`. The first part of the columns describe the subgroup intervals used
#'   for the biomarker variable, including where the center of the intervals are and their bounds. The
#'   second part of the columns contain the estimates for the treatment arm comparison.
#'
#' @note For the default degree 0 the `biomarker` variable is not included in the model.
#'
#' @seealso [control_step()] and [control_coxph()] for the available customization options.
#'
#' @examples
#' # Testing dataset with just two treatment arms.
#' library(dplyr)
#'
#' adtte_f <- tern_ex_adtte %>%
#'   filter(
#'     PARAMCD == "OS",
#'     ARM %in% c("B: Placebo", "A: Drug X")
#'   ) %>%
#'   mutate(
#'     # Reorder levels of ARM to display reference arm before treatment arm.
#'     ARM = droplevels(forcats::fct_relevel(ARM, "B: Placebo")),
#'     is_event = CNSR == 0
#'   )
#' labels <- c("ARM" = "Treatment Arm", "is_event" = "Event Flag")
#' formatters::var_labels(adtte_f)[names(labels)] <- labels
#'
#' variables <- list(
#'   arm = "ARM",
#'   biomarker = "BMRKR1",
#'   covariates = c("AGE", "BMRKR2"),
#'   event = "is_event",
#'   time = "AVAL"
#' )
#'
#' # Fit default STEP models: Here a constant treatment effect is estimated in each subgroup.
#' step_matrix <- fit_survival_step(
#'   variables = variables,
#'   data = adtte_f
#' )
#' dim(step_matrix)
#' head(step_matrix)
#'
#' # Specify different polynomial degree for the biomarker interaction to use more flexible local
#' # models. Or specify different Cox regression options.
#' step_matrix2 <- fit_survival_step(
#'   variables = variables,
#'   data = adtte_f,
#'   control = c(control_coxph(conf_level = 0.9), control_step(degree = 2))
#' )
#'
#' # Use a global model with cubic interaction and only 5 points.
#' step_matrix3 <- fit_survival_step(
#'   variables = variables,
#'   data = adtte_f,
#'   control = c(control_coxph(), control_step(bandwidth = NULL, degree = 3, num_points = 5L))
#' )
#'
#' @export
fit_survival_step <- function(variables,
                              data,
                              control = c(control_step(), control_coxph())) {
  checkmate::assert_list(control)
  assert_df_with_variables(data, variables)
  data <- data[!is.na(data[[variables$biomarker]]), ]
  window_sel <- h_step_window(x = data[[variables$biomarker]], control = control)
  interval_center <- window_sel$interval[, "Interval Center"]
  form <- h_step_survival_formula(variables = variables, control = control)
  estimates <- if (is.null(control$bandwidth)) {
    h_step_survival_est(
      formula = form,
      data = data,
      variables = variables,
      x = interval_center,
      control = control
    )
  } else {
    tmp <- mapply(
      FUN = h_step_survival_est,
      x = interval_center,
      subset = as.list(as.data.frame(window_sel$sel)),
      MoreArgs = list(
        formula = form,
        data = data,
        variables = variables,
        control = control
      )
    )
    # Maybe we find a more elegant solution than this.
    rownames(tmp) <- c("n", "events", "loghr", "se", "ci_lower", "ci_upper")
    t(tmp)
  }
  result <- cbind(window_sel$interval, estimates)
  structure(
    result,
    class = c("step", "matrix"),
    variables = variables,
    control = control
  )
}

#' Occurrence table sorting
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Functions to score occurrence table subtables and rows which can be used in the
#' sorting of occurrence tables.
#'
#' @name score_occurrences
NULL

#' @describeIn score_occurrences Scoring function which sums the counts across all
#'   columns. It will fail if anything else but counts are used.
#'
#' @inheritParams rtables_access
#'
#' @return
#' * `score_occurrences()` returns the sum of counts across all columns of a table row.
#'
#' @seealso [h_row_first_values()]
#'
#' @examples
#' lyt <- basic_table() %>%
#'   split_cols_by("ARM") %>%
#'   add_colcounts() %>%
#'   analyze_num_patients(
#'     vars = "USUBJID",
#'     .stats = c("unique"),
#'     .labels = c("Total number of patients with at least one event")
#'   ) %>%
#'   split_rows_by("AEBODSYS", child_labels = "visible", nested = FALSE) %>%
#'   summarize_num_patients(
#'     var = "USUBJID",
#'     .stats = c("unique", "nonunique"),
#'     .labels = c(
#'       "Total number of patients with at least one event",
#'       "Total number of events"
#'     )
#'   ) %>%
#'   count_occurrences(vars = "AEDECOD")
#'
#' tbl <- build_table(lyt, tern_ex_adae, alt_counts_df = tern_ex_adsl) %>%
#'   prune_table()
#'
#' tbl_sorted <- tbl %>%
#'   sort_at_path(path = c("AEBODSYS", "*", "AEDECOD"), scorefun = score_occurrences)
#'
#' tbl_sorted
#'
#' @export
score_occurrences <- function(table_row) {
  row_counts <- h_row_counts(table_row)
  sum(row_counts)
}

#' @describeIn score_occurrences Scoring functions can be produced by this constructor to only include
#'   specific columns in the scoring. See [h_row_counts()] for further information.
#'
#' @inheritParams has_count_in_cols
#'
#' @return
#' * `score_occurrences_cols()` returns a function that sums counts across all specified columns
#'   of a table row.
#'
#' @seealso [h_row_counts()]
#'
#' @examples
#' score_cols_a_and_b <- score_occurrences_cols(col_names = c("A: Drug X", "B: Placebo"))
#'
#' # Note that this here just sorts the AEDECOD inside the AEBODSYS. The AEBODSYS are not sorted.
#' # That would require a second pass of `sort_at_path`.
#' tbl_sorted <- tbl %>%
#'   sort_at_path(path = c("AEBODSYS", "*", "AEDECOD"), scorefun = score_cols_a_and_b)
#'
#' tbl_sorted
#'
#' @export
score_occurrences_cols <- function(...) {
  function(table_row) {
    row_counts <- h_row_counts(table_row, ...)
    sum(row_counts)
  }
}

#' @describeIn score_occurrences Scoring functions produced by this constructor can be used on
#'   subtables: They sum up all specified column counts in the subtable. This is useful when
#'   there is no available content row summing up these counts.
#'
#' @return
#' * `score_occurrences_subtable()` returns a function that sums counts in each subtable
#'   across all specified columns.
#'
#' @examples
#' score_subtable_all <- score_occurrences_subtable(col_names = names(tbl))
#'
#' # Note that this code just sorts the AEBODSYS, not the AEDECOD within AEBODSYS. That
#' # would require a second pass of `sort_at_path`.
#' tbl_sorted <- tbl %>%
#'   sort_at_path(path = c("AEBODSYS"), scorefun = score_subtable_all, decreasing = FALSE)
#'
#' tbl_sorted
#'
#' @export
score_occurrences_subtable <- function(...) {
  score_table_row <- score_occurrences_cols(...)
  function(table_tree) {
    table_rows <- collect_leaves(table_tree)
    counts <- vapply(table_rows, score_table_row, numeric(1))
    sum(counts)
  }
}

#' @describeIn score_occurrences Produces a score function for sorting table by summing the first content row in
#'   specified columns. Note that this is extending [rtables::cont_n_onecol()] and [rtables::cont_n_allcols()].
#'
#' @return
#' * `score_occurrences_cont_cols()` returns a function that sums counts in the first content row in
#'   specified columns.
#'
#' @export
score_occurrences_cont_cols <- function(...) {
  score_table_row <- score_occurrences_cols(...)
  function(table_tree) {
    if (inherits(table_tree, "ContentRow")) {
      return(NA)
    }
    content_row <- h_content_first_row(table_tree)
    score_table_row(content_row)
  }
}

#' Helper function to create a map data frame for `trim_levels_to_map()`
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Helper function to create a map data frame from the input dataset, which can be used as an argument in the
#' `trim_levels_to_map` split function. Based on different method, the map is constructed differently.
#'
#' @inheritParams argument_convention
#' @param abnormal (named `list`)\cr identifying the abnormal range level(s) in `df`. Based on the levels of
#'   abnormality of the input dataset, it can be something like `list(Low = "LOW LOW", High = "HIGH HIGH")` or
#'   `abnormal = list(Low = "LOW", High = "HIGH"))`
#' @param method (`string`)\cr indicates how the returned map will be constructed. Can be `"default"` or `"range"`.
#'
#' @return A map `data.frame`.
#'
#' @note If method is `"default"`, the returned map will only have the abnormal directions that are observed in the
#'   `df`, and records with all normal values will be excluded to avoid error in creating layout. If method is
#'   `"range"`, the returned map will be based on the rule that at least one observation with low range > 0
#'   for low direction and at least one observation with high range is not missing for high direction.
#'
#' @examples
#' adlb <- df_explicit_na(tern_ex_adlb)
#'
#' h_map_for_count_abnormal(
#'   df = adlb,
#'   variables = list(anl = "ANRIND", split_rows = c("LBCAT", "PARAM")),
#'   abnormal = list(low = c("LOW"), high = c("HIGH")),
#'   method = "default",
#'   na_str = "<Missing>"
#' )
#'
#' df <- data.frame(
#'   USUBJID = c(rep("1", 4), rep("2", 4), rep("3", 4)),
#'   AVISIT = c(
#'     rep("WEEK 1", 2),
#'     rep("WEEK 2", 2),
#'     rep("WEEK 1", 2),
#'     rep("WEEK 2", 2),
#'     rep("WEEK 1", 2),
#'     rep("WEEK 2", 2)
#'   ),
#'   PARAM = rep(c("ALT", "CPR"), 6),
#'   ANRIND = c(
#'     "NORMAL", "NORMAL", "LOW",
#'     "HIGH", "LOW", "LOW", "HIGH", "HIGH", rep("NORMAL", 4)
#'   ),
#'   ANRLO = rep(5, 12),
#'   ANRHI = rep(20, 12)
#' )
#' df$ANRIND <- factor(df$ANRIND, levels = c("LOW", "HIGH", "NORMAL"))
#' h_map_for_count_abnormal(
#'   df = df,
#'   variables = list(
#'     anl = "ANRIND",
#'     split_rows = c("PARAM"),
#'     range_low = "ANRLO",
#'     range_high = "ANRHI"
#'   ),
#'   abnormal = list(low = c("LOW"), high = c("HIGH")),
#'   method = "range",
#'   na_str = "<Missing>"
#' )
#'
#' @export
h_map_for_count_abnormal <- function(df,
                                     variables = list(
                                       anl = "ANRIND",
                                       split_rows = c("PARAM"),
                                       range_low = "ANRLO",
                                       range_high = "ANRHI"
                                     ),
                                     abnormal = list(low = c("LOW", "LOW LOW"), high = c("HIGH", "HIGH HIGH")),
                                     method = c("default", "range"),
                                     na_str = "<Missing>") {
  method <- match.arg(method)
  checkmate::assert_subset(c("anl", "split_rows"), names(variables))
  checkmate::assert_false(anyNA(df[variables$split_rows]))
  assert_df_with_variables(df,
    variables = list(anl = variables$anl, split_rows = variables$split_rows),
    na_level = na_str
  )
  assert_df_with_factors(df, list(val = variables$anl))
  assert_valid_factor(df[[variables$anl]], any.missing = FALSE)
  assert_list_of_variables(variables)
  checkmate::assert_list(abnormal, types = "character", len = 2)

  # Drop usued levels from df as they are not supposed to be in the final map
  df <- droplevels(df)

  normal_value <- setdiff(levels(df[[variables$anl]]), unlist(abnormal))

  # Based on the understanding of clinical data, there should only be one level of normal which is "NORMAL"
  checkmate::assert_vector(normal_value, len = 1)

  # Default method will only have what is observed in the df, and records with all normal values will be excluded to
  # avoid error in layout building.
  if (method == "default") {
    df_abnormal <- subset(df, df[[variables$anl]] %in% unlist(abnormal))
    map <- unique(df_abnormal[c(variables$split_rows, variables$anl)])
    map_normal <- unique(subset(map, select = variables$split_rows))
    map_normal[[variables$anl]] <- normal_value
    map <- rbind(map, map_normal)
  } else if (method == "range") {
    # range method follows the rule that at least one observation with ANRLO > 0 for low
    # direction and at least one observation with ANRHI is not missing for high direction.
    checkmate::assert_subset(c("range_low", "range_high"), names(variables))
    checkmate::assert_subset(c("LOW", "HIGH"), toupper(names(abnormal)))

    assert_df_with_variables(df,
      variables = list(
        range_low = variables$range_low,
        range_high = variables$range_high
      )
    )

    # Define low direction of map
    df_low <- subset(df, df[[variables$range_low]] > 0)
    map_low <- unique(df_low[variables$split_rows])
    low_levels <- unname(unlist(abnormal[toupper(names(abnormal)) == "LOW"]))
    low_levels_df <- as.data.frame(low_levels)
    colnames(low_levels_df) <- variables$anl
    low_levels_df <- do.call("rbind", replicate(nrow(map_low), low_levels_df, simplify = FALSE))
    rownames(map_low) <- NULL # Just to avoid strange row index in case upstream functions changed
    map_low <- map_low[rep(seq_len(nrow(map_low)), each = length(low_levels)), , drop = FALSE]
    map_low <- cbind(map_low, low_levels_df)

    # Define high direction of map
    df_high <- subset(df, df[[variables$range_high]] != na_str | !is.na(df[[variables$range_high]]))
    map_high <- unique(df_high[variables$split_rows])
    high_levels <- unname(unlist(abnormal[toupper(names(abnormal)) == "HIGH"]))
    high_levels_df <- as.data.frame(high_levels)
    colnames(high_levels_df) <- variables$anl
    high_levels_df <- do.call("rbind", replicate(nrow(map_high), high_levels_df, simplify = FALSE))
    rownames(map_high) <- NULL
    map_high <- map_high[rep(seq_len(nrow(map_high)), each = length(high_levels)), , drop = FALSE]
    map_high <- cbind(map_high, high_levels_df)

    # Define normal of map
    map_normal <- unique(rbind(map_low, map_high)[variables$split_rows])
    map_normal[variables$anl] <- normal_value

    map <- rbind(map_low, map_high, map_normal)
  }

  # map should be all characters
  map <- data.frame(lapply(map, as.character), stringsAsFactors = FALSE)

  # sort the map final output by split_rows variables
  for (i in rev(seq_len(length(variables$split_rows)))) {
    map <- map[order(map[[i]]), ]
  }
  map
}

#' Subgroup treatment effect pattern (STEP) fit for binary (response) outcome
#'
#' @description `r lifecycle::badge("stable")`
#'
#' This fits the Subgroup Treatment Effect Pattern logistic regression models for a binary
#' (response) outcome. The treatment arm variable must have exactly 2 levels,
#' where the first one is taken as reference and the estimated odds ratios are
#' for the comparison of the second level vs. the first one.
#'
#' The (conditional) logistic regression model which is fit is:
#'
#' `response ~ arm * poly(biomarker, degree) + covariates + strata(strata)`
#'
#' where `degree` is specified by `control_step()`.
#'
#' @inheritParams argument_convention
#' @param variables (named `list` of `character`)\cr list of analysis variables:
#'   needs `response`, `arm`, `biomarker`, and optional `covariates` and `strata`.
#' @param control (named `list`)\cr combined control list from [control_step()]
#'   and [control_logistic()].
#'
#' @return A matrix of class `step`. The first part of the columns describe the
#'   subgroup intervals used for the biomarker variable, including where the
#'   center of the intervals are and their bounds. The second part of the
#'   columns contain the estimates for the treatment arm comparison.
#'
#' @note For the default degree 0 the `biomarker` variable is not included in the model.
#'
#' @seealso [control_step()] and [control_logistic()] for the available
#'   customization options.
#'
#' @examples
#' # Testing dataset with just two treatment arms.
#' library(survival)
#' library(dplyr)
#'
#' adrs_f <- tern_ex_adrs %>%
#'   filter(
#'     PARAMCD == "BESRSPI",
#'     ARM %in% c("B: Placebo", "A: Drug X")
#'   ) %>%
#'   mutate(
#'     # Reorder levels of ARM to have Placebo as reference arm for Odds Ratio calculations.
#'     ARM = droplevels(forcats::fct_relevel(ARM, "B: Placebo")),
#'     RSP = case_when(AVALC %in% c("PR", "CR") ~ 1, TRUE ~ 0),
#'     SEX = factor(SEX)
#'   )
#'
#' variables <- list(
#'   arm = "ARM",
#'   biomarker = "BMRKR1",
#'   covariates = "AGE",
#'   response = "RSP"
#' )
#'
#' # Fit default STEP models: Here a constant treatment effect is estimated in each subgroup.
#' # We use a large enough bandwidth to avoid too small subgroups and linear separation in those.
#' step_matrix <- fit_rsp_step(
#'   variables = variables,
#'   data = adrs_f,
#'   control = c(control_logistic(), control_step(bandwidth = 0.9))
#' )
#' dim(step_matrix)
#' head(step_matrix)
#'
#' # Specify different polynomial degree for the biomarker interaction to use more flexible local
#' # models. Or specify different logistic regression options, including confidence level.
#' step_matrix2 <- fit_rsp_step(
#'   variables = variables,
#'   data = adrs_f,
#'   control = c(control_logistic(conf_level = 0.9), control_step(bandwidth = NULL, degree = 1))
#' )
#'
#' # Use a global constant model. This is helpful as a reference for the subgroup models.
#' step_matrix3 <- fit_rsp_step(
#'   variables = variables,
#'   data = adrs_f,
#'   control = c(control_logistic(), control_step(bandwidth = NULL, num_points = 2L))
#' )
#'
#' # It is also possible to use strata, i.e. use conditional logistic regression models.
#' variables2 <- list(
#'   arm = "ARM",
#'   biomarker = "BMRKR1",
#'   covariates = "AGE",
#'   response = "RSP",
#'   strata = c("STRATA1", "STRATA2")
#' )
#'
#' step_matrix4 <- fit_rsp_step(
#'   variables = variables2,
#'   data = adrs_f,
#'   control = c(control_logistic(), control_step(bandwidth = NULL))
#' )
#'
#' @export
fit_rsp_step <- function(variables,
                         data,
                         control = c(control_step(), control_logistic())) {
  assert_df_with_variables(data, variables)
  checkmate::assert_list(control, names = "named")
  data <- data[!is.na(data[[variables$biomarker]]), ]
  window_sel <- h_step_window(x = data[[variables$biomarker]], control = control)
  interval_center <- window_sel$interval[, "Interval Center"]
  form <- h_step_rsp_formula(variables = variables, control = control)
  estimates <- if (is.null(control$bandwidth)) {
    h_step_rsp_est(
      formula = form,
      data = data,
      variables = variables,
      x = interval_center,
      control = control
    )
  } else {
    tmp <- mapply(
      FUN = h_step_rsp_est,
      x = interval_center,
      subset = as.list(as.data.frame(window_sel$sel)),
      MoreArgs = list(
        formula = form,
        data = data,
        variables = variables,
        control = control
      )
    )
    # Maybe we find a more elegant solution than this.
    rownames(tmp) <- c("n", "logor", "se", "ci_lower", "ci_upper")
    t(tmp)
  }
  result <- cbind(window_sel$interval, estimates)
  structure(
    result,
    class = c("step", "matrix"),
    variables = variables,
    control = control
  )
}

#' Helper functions for tabulating biomarker effects on survival by subgroup
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Helper functions which are documented here separately to not confuse the user
#' when reading about the user-facing functions.
#'
#' @inheritParams survival_biomarkers_subgroups
#' @inheritParams argument_convention
#' @inheritParams fit_coxreg_multivar
#'
#' @examples
#' library(dplyr)
#' library(forcats)
#'
#' adtte <- tern_ex_adtte
#'
#' # Save variable labels before data processing steps.
#' adtte_labels <- formatters::var_labels(adtte, fill = FALSE)
#'
#' adtte_f <- adtte %>%
#'   filter(PARAMCD == "OS") %>%
#'   mutate(
#'     AVALU = as.character(AVALU),
#'     is_event = CNSR == 0
#'   )
#' labels <- c("AVALU" = adtte_labels[["AVALU"]], "is_event" = "Event Flag")
#' formatters::var_labels(adtte_f)[names(labels)] <- labels
#'
#' @name h_survival_biomarkers_subgroups
NULL

#' @describeIn h_survival_biomarkers_subgroups Helps with converting the "survival" function variable list
#'   to the "Cox regression" variable list. The reason is that currently there is an inconsistency between the variable
#'   names accepted by `extract_survival_subgroups()` and `fit_coxreg_multivar()`.
#'
#' @param biomarker (`string`)\cr the name of the biomarker variable.
#'
#' @return
#' * `h_surv_to_coxreg_variables()` returns a named `list` of elements `time`, `event`, `arm`,
#'   `covariates`, and `strata`.
#'
#' @examples
#' # This is how the variable list is converted internally.
#' h_surv_to_coxreg_variables(
#'   variables = list(
#'     tte = "AVAL",
#'     is_event = "EVNT",
#'     covariates = c("A", "B"),
#'     strata = "D"
#'   ),
#'   biomarker = "AGE"
#' )
#'
#' @export
h_surv_to_coxreg_variables <- function(variables, biomarker) {
  checkmate::assert_list(variables)
  checkmate::assert_string(variables$tte)
  checkmate::assert_string(variables$is_event)
  checkmate::assert_string(biomarker)
  list(
    time = variables$tte,
    event = variables$is_event,
    arm = biomarker,
    covariates = variables$covariates,
    strata = variables$strata
  )
}

#' @describeIn h_survival_biomarkers_subgroups Prepares estimates for number of events, patients and median survival
#'   times, as well as hazard ratio estimates, confidence intervals and p-values, for multiple biomarkers
#'   in a given single data set.
#'   `variables` corresponds to names of variables found in `data`, passed as a named list and requires elements
#'   `tte`, `is_event`, `biomarkers` (vector of continuous biomarker variables) and optionally `subgroups` and `strata`.
#'
#' @return
#' * `h_coxreg_mult_cont_df()` returns a `data.frame` containing estimates and statistics for the selected biomarkers.
#'
#' @examples
#' # For a single population, estimate separately the effects
#' # of two biomarkers.
#' df <- h_coxreg_mult_cont_df(
#'   variables = list(
#'     tte = "AVAL",
#'     is_event = "is_event",
#'     biomarkers = c("BMRKR1", "AGE"),
#'     covariates = "SEX",
#'     strata = c("STRATA1", "STRATA2")
#'   ),
#'   data = adtte_f
#' )
#' df
#'
#' # If the data set is empty, still the corresponding rows with missings are returned.
#' h_coxreg_mult_cont_df(
#'   variables = list(
#'     tte = "AVAL",
#'     is_event = "is_event",
#'     biomarkers = c("BMRKR1", "AGE"),
#'     covariates = "REGION1",
#'     strata = c("STRATA1", "STRATA2")
#'   ),
#'   data = adtte_f[NULL, ]
#' )
#'
#' @export
h_coxreg_mult_cont_df <- function(variables,
                                  data,
                                  control = control_coxreg()) {
  if ("strat" %in% names(variables)) {
    warning(
      "Warning: the `strat` element name of the `variables` list argument to `h_coxreg_mult_cont_df() ",
      "was deprecated in tern 0.9.4.\n  ",
      "Please use the name `strata` instead of `strat` in the `variables` argument."
    )
    variables[["strata"]] <- variables[["strat"]]
  }

  assert_df_with_variables(data, variables)
  checkmate::assert_list(control, names = "named")
  checkmate::assert_character(variables$biomarkers, min.len = 1, any.missing = FALSE)
  conf_level <- control[["conf_level"]]
  pval_label <- paste0(
    # the regex capitalizes the first letter of the string / senetence.
    "p-value (", gsub("(^[a-z])", "\\U\\1", trimws(control[["pval_method"]]), perl = TRUE), ")"
  )
  # If there is any data, run model, otherwise return empty results.
  if (nrow(data) > 0) {
    bm_cols <- match(variables$biomarkers, names(data))
    l_result <- lapply(variables$biomarkers, function(bm) {
      coxreg_list <- fit_coxreg_multivar(
        variables = h_surv_to_coxreg_variables(variables, bm),
        data = data,
        control = control
      )
      result <- do.call(
        h_coxreg_multivar_extract,
        c(list(var = bm), coxreg_list[c("mod", "data", "control")])
      )
      data_fit <- as.data.frame(as.matrix(coxreg_list$mod$y))
      data_fit$status <- as.logical(data_fit$status)
      median <- s_surv_time(
        df = data_fit,
        .var = "time",
        is_event = "status"
      )$median
      data.frame(
        # Dummy column needed downstream to create a nested header.
        biomarker = bm,
        biomarker_label = formatters::var_labels(data[bm], fill = TRUE),
        n_tot = coxreg_list$mod$n,
        n_tot_events = coxreg_list$mod$nevent,
        median = as.numeric(median),
        result[1L, c("hr", "lcl", "ucl")],
        conf_level = conf_level,
        pval = result[1L, "pval"],
        pval_label = pval_label,
        stringsAsFactors = FALSE
      )
    })
    do.call(rbind, args = c(l_result, make.row.names = FALSE))
  } else {
    data.frame(
      biomarker = variables$biomarkers,
      biomarker_label = formatters::var_labels(data[variables$biomarkers], fill = TRUE),
      n_tot = 0L,
      n_tot_events = 0L,
      median = NA,
      hr = NA,
      lcl = NA,
      ucl = NA,
      conf_level = conf_level,
      pval = NA,
      pval_label = pval_label,
      row.names = seq_along(variables$biomarkers),
      stringsAsFactors = FALSE
    )
  }
}

#' @describeIn h_survival_biomarkers_subgroups Prepares a single sub-table given a `df_sub` containing
#'   the results for a single biomarker.
#'
#' @param df (`data.frame`)\cr results for a single biomarker, as part of what is
#'   returned by [extract_survival_biomarkers()] (it needs a couple of columns which are
#'   added by that high-level function relative to what is returned by [h_coxreg_mult_cont_df()],
#'   see the example).
#'
#' @return
#' * `h_tab_surv_one_biomarker()` returns an `rtables` table object with the given statistics arranged in columns.
#'
#' @examples
#' # Starting from above `df`, zoom in on one biomarker and add required columns.
#' df1 <- df[1, ]
#' df1$subgroup <- "All patients"
#' df1$row_type <- "content"
#' df1$var <- "ALL"
#' df1$var_label <- "All patients"
#' h_tab_surv_one_biomarker(
#'   df1,
#'   vars = c("n_tot", "n_tot_events", "median", "hr", "ci", "pval"),
#'   time_unit = "days"
#' )
#'
#' @export
h_tab_surv_one_biomarker <- function(df,
                                     vars,
                                     time_unit,
                                     na_str = default_na_str(),
                                     .indent_mods = 0L,
                                     ...) {
  afuns <- a_survival_subgroups(na_str = na_str)[vars]
  colvars <- d_survival_subgroups_colvars(
    vars,
    conf_level = df$conf_level[1],
    method = df$pval_label[1],
    time_unit = time_unit
  )
  h_tab_one_biomarker(
    df = df,
    afuns = afuns,
    colvars = colvars,
    na_str = na_str,
    .indent_mods = .indent_mods,
    ...
  )
}

#' Control function for descriptive statistics
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Sets a list of parameters for summaries of descriptive statistics. Typically used internally to specify
#' details for [s_summary()]. This function family is mainly used by [analyze_vars()].
#'
#' @inheritParams argument_convention
#' @param quantiles (`numeric(2)`)\cr vector of length two to specify the quantiles to calculate.
#' @param quantile_type (`numeric(1)`)\cr number between 1 and 9 selecting quantile algorithms to be used.
#'   Default is set to 2 as this matches the default quantile algorithm in SAS `proc univariate` set by `QNTLDEF=5`.
#'   This differs from R's default. See more about `type` in [stats::quantile()].
#' @param test_mean (`numeric(1)`)\cr number to test against the mean under the null hypothesis when calculating
#'   p-value.
#'
#' @return A list of components with the same names as the arguments.
#'
#' @export
control_analyze_vars <- function(conf_level = 0.95,
                                 quantiles = c(0.25, 0.75),
                                 quantile_type = 2,
                                 test_mean = 0) {
  checkmate::assert_vector(quantiles, len = 2)
  checkmate::assert_int(quantile_type, lower = 1, upper = 9)
  checkmate::assert_numeric(test_mean)
  lapply(quantiles, assert_proportion_value)
  assert_proportion_value(conf_level)
  list(conf_level = conf_level, quantiles = quantiles, quantile_type = quantile_type, test_mean = test_mean)
}

#' Analyze variables
#'
#' @description `r lifecycle::badge("stable")`
#'
#' The analyze function [analyze_vars()] creates a layout element to summarize one or more variables, using the S3
#' generic function [s_summary()] to calculate a list of summary statistics. A list of all available statistics for
#' numeric variables can be viewed by running `get_stats("analyze_vars_numeric")` and for non-numeric variables by
#' running `get_stats("analyze_vars_counts")`. Use the `.stats` parameter to specify the statistics to include in your
#' output summary table.
#'
#' @details
#' **Automatic digit formatting:** The number of digits to display can be automatically determined from the analyzed
#' variable(s) (`vars`) for certain statistics by setting the statistic format to `"auto"` in `.formats`.
#' This utilizes the [format_auto()] formatting function. Note that only data for the current row & variable (for all
#' columns) will be considered (`.df_row[[.var]]`, see [`rtables::additional_fun_params`]) and not the whole dataset.
#'
#' @inheritParams argument_convention
#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("analyze_vars_numeric")` to see
#'   statistics available for numeric variables, and `get_stats("analyze_vars_counts")` for statistics available
#'   for non-numeric variables.
#'
#' @name analyze_variables
#' @order 1
NULL

#' @describeIn analyze_variables S3 generic function to produces a variable summary.
#'
#' @return
#' * `s_summary()` returns different statistics depending on the class of `x`.
#'
#' @export
s_summary <- function(x,
                      na.rm = TRUE, # nolint
                      denom,
                      .N_row, # nolint
                      .N_col, # nolint
                      .var,
                      ...) {
  checkmate::assert_flag(na.rm)
  UseMethod("s_summary", x)
}

#' @describeIn analyze_variables Method for `numeric` class.
#'
#' @param control (`list`)\cr parameters for descriptive statistics details, specified by using
#'   the helper function [control_analyze_vars()]. Some possible parameter options are:
#'   * `conf_level` (`proportion`)\cr confidence level of the interval for mean and median.
#'   * `quantiles` (`numeric(2)`)\cr vector of length two to specify the quantiles.
#'   * `quantile_type` (`numeric(1)`)\cr between 1 and 9 selecting quantile algorithms to be used.
#'     See more about `type` in [stats::quantile()].
#'   * `test_mean` (`numeric(1)`)\cr value to test against the mean under the null hypothesis when calculating p-value.
#'
#' @return
#'   * If `x` is of class `numeric`, returns a `list` with the following named `numeric` items:
#'     * `n`: The [length()] of `x`.
#'     * `sum`: The [sum()] of `x`.
#'     * `mean`: The [mean()] of `x`.
#'     * `sd`: The [stats::sd()] of `x`.
#'     * `se`: The standard error of `x` mean, i.e.: (`sd(x) / sqrt(length(x))`).
#'     * `mean_sd`: The [mean()] and [stats::sd()] of `x`.
#'     * `mean_se`: The [mean()] of `x` and its standard error (see above).
#'     * `mean_ci`: The CI for the mean of `x` (from [stat_mean_ci()]).
#'     * `mean_sei`: The SE interval for the mean of `x`, i.e.: ([mean()] -/+ [stats::sd()] / [sqrt()]).
#'     * `mean_sdi`: The SD interval for the mean of `x`, i.e.: ([mean()] -/+ [stats::sd()]).
#'     * `mean_pval`: The two-sided p-value of the mean of `x` (from [stat_mean_pval()]).
#'     * `median`: The [stats::median()] of `x`.
#'     * `mad`: The median absolute deviation of `x`, i.e.: ([stats::median()] of `xc`,
#'       where `xc` = `x` - [stats::median()]).
#'     * `median_ci`: The CI for the median of `x` (from [stat_median_ci()]).
#'     * `quantiles`: Two sample quantiles of `x` (from [stats::quantile()]).
#'     * `iqr`: The [stats::IQR()] of `x`.
#'     * `range`: The [range_noinf()] of `x`.
#'     * `min`: The [max()] of `x`.
#'     * `max`: The [min()] of `x`.
#'     * `median_range`: The [median()] and [range_noinf()] of `x`.
#'     * `cv`: The coefficient of variation of `x`, i.e.: ([stats::sd()] / [mean()] * 100).
#'     * `geom_mean`: The geometric mean of `x`, i.e.: (`exp(mean(log(x)))`).
#'     * `geom_cv`: The geometric coefficient of variation of `x`, i.e.: (`sqrt(exp(sd(log(x)) ^ 2) - 1) * 100`).
#'
#' @note
#' * If `x` is an empty vector, `NA` is returned. This is the expected feature so as to return `rcell` content in
#'   `rtables` when the intersection of a column and a row delimits an empty data selection.
#' * When the `mean` function is applied to an empty vector, `NA` will be returned instead of `NaN`, the latter
#'   being standard behavior in R.
#'
#' @method s_summary numeric
#'
#' @examples
#' # `s_summary.numeric`
#'
#' ## Basic usage: empty numeric returns NA-filled items.
#' s_summary(numeric())
#'
#' ## Management of NA values.
#' x <- c(NA_real_, 1)
#' s_summary(x, na.rm = TRUE)
#' s_summary(x, na.rm = FALSE)
#'
#' x <- c(NA_real_, 1, 2)
#' s_summary(x, stats = NULL)
#'
#' ## Benefits in `rtables` contructions:
#' dta_test <- data.frame(
#'   Group = rep(LETTERS[1:3], each = 2),
#'   sub_group = rep(letters[1:2], each = 3),
#'   x = 1:6
#' )
#'
#' ## The summary obtained in with `rtables`:
#' basic_table() %>%
#'   split_cols_by(var = "Group") %>%
#'   split_rows_by(var = "sub_group") %>%
#'   analyze(vars = "x", afun = s_summary) %>%
#'   build_table(df = dta_test)
#'
#' ## By comparison with `lapply`:
#' X <- split(dta_test, f = with(dta_test, interaction(Group, sub_group)))
#' lapply(X, function(x) s_summary(x$x))
#'
#' @export
s_summary.numeric <- function(x,
                              na.rm = TRUE, # nolint
                              denom,
                              .N_row, # nolint
                              .N_col, # nolint
                              .var,
                              control = control_analyze_vars(),
                              ...) {
  checkmate::assert_numeric(x)

  if (na.rm) {
    x <- x[!is.na(x)]
  }

  y <- list()

  y$n <- c("n" = length(x))

  y$sum <- c("sum" = ifelse(length(x) == 0, NA_real_, sum(x, na.rm = FALSE)))

  y$mean <- c("mean" = ifelse(length(x) == 0, NA_real_, mean(x, na.rm = FALSE)))

  y$sd <- c("sd" = stats::sd(x, na.rm = FALSE))

  y$se <- c("se" = stats::sd(x, na.rm = FALSE) / sqrt(length(stats::na.omit(x))))

  y$mean_sd <- c(y$mean, "sd" = stats::sd(x, na.rm = FALSE))

  y$mean_se <- c(y$mean, y$se)

  mean_ci <- stat_mean_ci(x, conf_level = control$conf_level, na.rm = FALSE, gg_helper = FALSE)
  y$mean_ci <- formatters::with_label(mean_ci, paste("Mean", f_conf_level(control$conf_level)))

  mean_sei <- y$mean[[1]] + c(-1, 1) * stats::sd(x, na.rm = FALSE) / sqrt(y$n)
  names(mean_sei) <- c("mean_sei_lwr", "mean_sei_upr")
  y$mean_sei <- formatters::with_label(mean_sei, "Mean -/+ 1xSE")

  mean_sdi <- y$mean[[1]] + c(-1, 1) * stats::sd(x, na.rm = FALSE)
  names(mean_sdi) <- c("mean_sdi_lwr", "mean_sdi_upr")
  y$mean_sdi <- formatters::with_label(mean_sdi, "Mean -/+ 1xSD")

  mean_pval <- stat_mean_pval(x, test_mean = control$test_mean, na.rm = FALSE, n_min = 2)
  y$mean_pval <- formatters::with_label(mean_pval, paste("Mean", f_pval(control$test_mean)))

  y$median <- c("median" = stats::median(x, na.rm = FALSE))

  y$mad <- c("mad" = stats::median(x - y$median, na.rm = FALSE))

  median_ci <- stat_median_ci(x, conf_level = control$conf_level, na.rm = FALSE, gg_helper = FALSE)
  y$median_ci <- formatters::with_label(median_ci, paste("Median", f_conf_level(control$conf_level)))

  q <- control$quantiles
  if (any(is.na(x))) {
    qnts <- rep(NA_real_, length(q))
  } else {
    qnts <- stats::quantile(x, probs = q, type = control$quantile_type, na.rm = FALSE)
  }
  names(qnts) <- paste("quantile", q, sep = "_")
  y$quantiles <- formatters::with_label(qnts, paste0(paste(paste0(q * 100, "%"), collapse = " and "), "-ile"))

  y$iqr <- c("iqr" = ifelse(
    any(is.na(x)),
    NA_real_,
    stats::IQR(x, na.rm = FALSE, type = control$quantile_type)
  ))

  y$range <- stats::setNames(range_noinf(x, na.rm = FALSE), c("min", "max"))
  y$min <- y$range[1]
  y$max <- y$range[2]

  y$median_range <- formatters::with_label(c(y$median, y$range), "Median (Min - Max)")

  y$cv <- c("cv" = unname(y$sd) / unname(y$mean) * 100)

  # Convert negative values to NA for log calculation.
  x_no_negative_vals <- x
  x_no_negative_vals[x_no_negative_vals <= 0] <- NA
  y$geom_mean <- c("geom_mean" = exp(mean(log(x_no_negative_vals), na.rm = FALSE)))
  geom_mean_ci <- stat_mean_ci(x, conf_level = control$conf_level, na.rm = FALSE, gg_helper = FALSE, geom_mean = TRUE)
  y$geom_mean_ci <- formatters::with_label(geom_mean_ci, paste("Geometric Mean", f_conf_level(control$conf_level)))

  y$geom_cv <- c("geom_cv" = sqrt(exp(stats::sd(log(x_no_negative_vals), na.rm = FALSE) ^ 2) - 1) * 100) # styler: off

  y
}

#' @describeIn analyze_variables Method for `factor` class.
#'
#' @param denom (`string`)\cr choice of denominator for factor proportions. Options are:
#'   * `n`: number of values in this row and column intersection.
#'   * `N_row`: total number of values in this row across columns.
#'   * `N_col`: total number of values in this column across rows.
#'
#' @return
#'   * If `x` is of class `factor` or converted from `character`, returns a `list` with named `numeric` items:
#'     * `n`: The [length()] of `x`.
#'     * `count`: A list with the number of cases for each level of the factor `x`.
#'     * `count_fraction`: Similar to `count` but also includes the proportion of cases for each level of the
#'       factor `x` relative to the denominator, or `NA` if the denominator is zero.
#'
#' @note
#' * If `x` is an empty `factor`, a list is still returned for `counts` with one element
#'   per factor level. If there are no levels in `x`, the function fails.
#' * If factor variables contain `NA`, these `NA` values are excluded by default. To include `NA` values
#'   set `na.rm = FALSE` and missing values will be displayed as an `NA` level. Alternatively, an explicit
#'   factor level can be defined for `NA` values during pre-processing via [df_explicit_na()] - the
#'   default `na_level` (`"<Missing>"`) will also be excluded when `na.rm` is set to `TRUE`.
#'
#' @method s_summary factor
#'
#' @examples
#' # `s_summary.factor`
#'
#' ## Basic usage:
#' s_summary(factor(c("a", "a", "b", "c", "a")))
#'
#' # Empty factor returns zero-filled items.
#' s_summary(factor(levels = c("a", "b", "c")))
#'
#' ## Management of NA values.
#' x <- factor(c(NA, "Female"))
#' x <- explicit_na(x)
#' s_summary(x, na.rm = TRUE)
#' s_summary(x, na.rm = FALSE)
#'
#' ## Different denominators.
#' x <- factor(c("a", "a", "b", "c", "a"))
#' s_summary(x, denom = "N_row", .N_row = 10L)
#' s_summary(x, denom = "N_col", .N_col = 20L)
#'
#' @export
s_summary.factor <- function(x,
                             na.rm = TRUE, # nolint
                             denom = c("n", "N_row", "N_col"),
                             .N_row, # nolint
                             .N_col, # nolint
                             ...) {
  assert_valid_factor(x)
  denom <- match.arg(denom)

  if (na.rm) {
    x <- x[!is.na(x)] %>% fct_discard("<Missing>")
  } else {
    x <- x %>% explicit_na(label = "NA")
  }

  y <- list()

  y$n <- length(x)

  y$count <- as.list(table(x, useNA = "ifany"))
  dn <- switch(denom,
    n = length(x),
    N_row = .N_row,
    N_col = .N_col
  )
  y$count_fraction <- lapply(
    y$count,
    function(x) {
      c(x, ifelse(dn > 0, x / dn, 0))
    }
  )
  y$fraction <- lapply(
    y$count,
    function(count) c("num" = count, "denom" = dn)
  )

  y$n_blq <- sum(grepl("BLQ|LTR|<[1-9]|<PCLLOQ", x))

  y
}

#' @describeIn analyze_variables Method for `character` class. This makes an automatic
#'   conversion to factor (with a warning) and then forwards to the method for factors.
#'
#' @param verbose (`flag`)\cr defaults to `TRUE`, which prints out warnings and messages. It is mainly used
#'   to print out information about factor casting.
#'
#' @note
#' * Automatic conversion of character to factor does not guarantee that the table
#'   can be generated correctly. In particular for sparse tables this very likely can fail.
#'   It is therefore better to always pre-process the dataset such that factors are manually
#'   created from character variables before passing the dataset to [rtables::build_table()].
#'
#' @method s_summary character
#'
#' @examples
#' # `s_summary.character`
#'
#' ## Basic usage:
#' s_summary(c("a", "a", "b", "c", "a"), .var = "x", verbose = FALSE)
#' s_summary(c("a", "a", "b", "c", "a", ""), .var = "x", na.rm = FALSE, verbose = FALSE)
#'
#' @export
s_summary.character <- function(x,
                                na.rm = TRUE, # nolint
                                denom = c("n", "N_row", "N_col"),
                                .N_row, # nolint
                                .N_col, # nolint
                                .var,
                                verbose = TRUE,
                                ...) {
  if (na.rm) {
    y <- as_factor_keep_attributes(x, verbose = verbose)
  } else {
    y <- as_factor_keep_attributes(x, verbose = verbose, na_level = "NA")
  }

  s_summary(
    x = y,
    na.rm = na.rm,
    denom = denom,
    .N_row = .N_row,
    .N_col = .N_col,
    ...
  )
}

#' @describeIn analyze_variables Method for `logical` class.
#'
#' @param denom (`string`)\cr choice of denominator for proportion. Options are:
#'   * `n`: number of values in this row and column intersection.
#'   * `N_row`: total number of values in this row across columns.
#'   * `N_col`: total number of values in this column across rows.
#'
#' @return
#'   * If `x` is of class `logical`, returns a `list` with named `numeric` items:
#'     * `n`: The [length()] of `x` (possibly after removing `NA`s).
#'     * `count`: Count of `TRUE` in `x`.
#'     * `count_fraction`: Count and proportion of `TRUE` in `x` relative to the denominator, or `NA` if the
#'       denominator is zero. Note that `NA`s in `x` are never counted or leading to `NA` here.
#'
#' @method s_summary logical
#'
#' @examples
#' # `s_summary.logical`
#'
#' ## Basic usage:
#' s_summary(c(TRUE, FALSE, TRUE, TRUE))
#'
#' # Empty factor returns zero-filled items.
#' s_summary(as.logical(c()))
#'
#' ## Management of NA values.
#' x <- c(NA, TRUE, FALSE)
#' s_summary(x, na.rm = TRUE)
#' s_summary(x, na.rm = FALSE)
#'
#' ## Different denominators.
#' x <- c(TRUE, FALSE, TRUE, TRUE)
#' s_summary(x, denom = "N_row", .N_row = 10L)
#' s_summary(x, denom = "N_col", .N_col = 20L)
#'
#' @export
s_summary.logical <- function(x,
                              na.rm = TRUE, # nolint
                              denom = c("n", "N_row", "N_col"),
                              .N_row, # nolint
                              .N_col, # nolint
                              ...) {
  denom <- match.arg(denom)
  if (na.rm) x <- x[!is.na(x)]
  y <- list()
  y$n <- length(x)
  count <- sum(x, na.rm = TRUE)
  dn <- switch(denom,
    n = length(x),
    N_row = .N_row,
    N_col = .N_col
  )
  y$count <- count
  y$count_fraction <- c(count, ifelse(dn > 0, count / dn, 0))
  y$n_blq <- 0L
  y
}

#' @describeIn analyze_variables Formatted analysis function which is used as `afun` in `analyze_vars()` and
#'   `compare_vars()` and as `cfun` in `summarize_colvars()`.
#'
#' @param compare (`flag`)\cr whether comparison statistics should be analyzed instead of summary statistics
#'   (`compare = TRUE` adds `pval` statistic comparing against reference group).
#'
#' @return
#' * `a_summary()` returns the corresponding list with formatted [rtables::CellValue()].
#'
#' @note
#' * To use for comparison (with additional p-value statistic), parameter `compare` must be set to `TRUE`.
#' * Ensure that either all `NA` values are converted to an explicit `NA` level or all `NA` values are left as is.
#'
#' @examples
#' a_summary(factor(c("a", "a", "b", "c", "a")), .N_row = 10, .N_col = 10)
#' a_summary(
#'   factor(c("a", "a", "b", "c", "a")),
#'   .ref_group = factor(c("a", "a", "b", "c")), compare = TRUE
#' )
#'
#' a_summary(c("A", "B", "A", "C"), .var = "x", .N_col = 10, .N_row = 10, verbose = FALSE)
#' a_summary(
#'   c("A", "B", "A", "C"),
#'   .ref_group = c("B", "A", "C"), .var = "x", compare = TRUE, verbose = FALSE
#' )
#'
#' a_summary(c(TRUE, FALSE, FALSE, TRUE, TRUE), .N_row = 10, .N_col = 10)
#' a_summary(
#'   c(TRUE, FALSE, FALSE, TRUE, TRUE),
#'   .ref_group = c(TRUE, FALSE), .in_ref_col = TRUE, compare = TRUE
#' )
#'
#' a_summary(rnorm(10), .N_col = 10, .N_row = 20, .var = "bla")
#' a_summary(rnorm(10, 5, 1), .ref_group = rnorm(20, -5, 1), .var = "bla", compare = TRUE)
#'
#' @export
a_summary <- function(x,
                      .N_col, # nolint
                      .N_row, # nolint
                      .var = NULL,
                      .df_row = NULL,
                      .ref_group = NULL,
                      .in_ref_col = FALSE,
                      compare = FALSE,
                      .stats = NULL,
                      .formats = NULL,
                      .labels = NULL,
                      .indent_mods = NULL,
                      na.rm = TRUE, # nolint
                      na_str = default_na_str(),
                      ...) {
  extra_args <- list(...)
  if (is.numeric(x)) {
    type <- "numeric"
    if (!is.null(.stats) && any(grepl("^pval", .stats))) {
      .stats[grepl("^pval", .stats)] <- "pval" # tmp fix xxx
    }
  } else {
    type <- "counts"
    if (!is.null(.stats) && any(grepl("^pval", .stats))) {
      .stats[grepl("^pval", .stats)] <- "pval_counts" # tmp fix xxx
    }
  }

  # If one col has NA vals, must add NA row to other cols (using placeholder lvl `fill-na-level`)
  if (any(is.na(.df_row[[.var]])) && !any(is.na(x)) && !na.rm) levels(x) <- c(levels(x), "fill-na-level")

  x_stats <- if (!compare) {
    s_summary(x = x, .N_col = .N_col, .N_row = .N_row, na.rm = na.rm, ...)
  } else {
    s_compare(
      x = x, .N_col = .N_col, .N_row = .N_row, na.rm = na.rm, .ref_group = .ref_group, .in_ref_col = .in_ref_col, ...
    )
  }

  # Fill in with formatting defaults if needed
  met_grp <- paste0(c("analyze_vars", type), collapse = "_")
  .stats <- get_stats(met_grp, stats_in = .stats, add_pval = compare)
  .formats <- get_formats_from_stats(.stats, .formats)
  .indent_mods <- get_indents_from_stats(.stats, .indent_mods)

  lbls <- get_labels_from_stats(.stats, .labels)
  # Check for custom labels from control_analyze_vars
  .labels <- if ("control" %in% names(extra_args)) {
    lbls %>% labels_use_control(extra_args[["control"]], .labels)
  } else {
    lbls
  }

  if ("count_fraction_fixed_dp" %in% .stats) x_stats[["count_fraction_fixed_dp"]] <- x_stats[["count_fraction"]]
  x_stats <- x_stats[.stats]

  if (is.factor(x) || is.character(x)) {
    # Ungroup statistics with values for each level of x
    x_ungrp <- ungroup_stats(x_stats, .formats, .labels, .indent_mods)
    x_stats <- x_ungrp[["x"]]
    .formats <- x_ungrp[[".formats"]]
    .labels <- gsub("fill-na-level", "NA", x_ungrp[[".labels"]])
    .indent_mods <- x_ungrp[[".indent_mods"]]
  }

  # Auto format handling
  .formats <- apply_auto_formatting(.formats, x_stats, .df_row, .var)

  in_rows(
    .list = x_stats,
    .formats = .formats,
    .names = .labels,
    .labels = .labels,
    .indent_mods = .indent_mods,
    .format_na_strs = na_str
  )
}

#' @describeIn analyze_variables Layout-creating function which can take statistics function arguments
#'   and additional format arguments. This function is a wrapper for [rtables::analyze()].
#'
#' @param ... arguments passed to `s_summary()`.
#' @param .indent_mods (named `integer`)\cr indent modifiers for the labels. Each element of the vector
#'   should be a name-value pair with name corresponding to a statistic specified in `.stats` and value the indentation
#'   for that statistic's row label.
#'
#' @return
#' * `analyze_vars()` returns a layout object suitable for passing to further layouting functions,
#'   or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
#'   the statistics from `s_summary()` to the table layout.
#'
#' @examples
#' ## Fabricated dataset.
#' dta_test <- data.frame(
#'   USUBJID = rep(1:6, each = 3),
#'   PARAMCD = rep("lab", 6 * 3),
#'   AVISIT  = rep(paste0("V", 1:3), 6),
#'   ARM     = rep(LETTERS[1:3], rep(6, 3)),
#'   AVAL    = c(9:1, rep(NA, 9))
#' )
#'
#' # `analyze_vars()` in `rtables` pipelines
#' ## Default output within a `rtables` pipeline.
#' l <- basic_table() %>%
#'   split_cols_by(var = "ARM") %>%
#'   split_rows_by(var = "AVISIT") %>%
#'   analyze_vars(vars = "AVAL")
#'
#' build_table(l, df = dta_test)
#'
#' ## Select and format statistics output.
#' l <- basic_table() %>%
#'   split_cols_by(var = "ARM") %>%
#'   split_rows_by(var = "AVISIT") %>%
#'   analyze_vars(
#'     vars = "AVAL",
#'     .stats = c("n", "mean_sd", "quantiles"),
#'     .formats = c("mean_sd" = "xx.x, xx.x"),
#'     .labels = c(n = "n", mean_sd = "Mean, SD", quantiles = c("Q1 - Q3"))
#'   )
#'
#' build_table(l, df = dta_test)
#'
#' ## Use arguments interpreted by `s_summary`.
#' l <- basic_table() %>%
#'   split_cols_by(var = "ARM") %>%
#'   split_rows_by(var = "AVISIT") %>%
#'   analyze_vars(vars = "AVAL", na.rm = FALSE)
#'
#' build_table(l, df = dta_test)
#'
#' ## Handle `NA` levels first when summarizing factors.
#' dta_test$AVISIT <- NA_character_
#' dta_test <- df_explicit_na(dta_test)
#' l <- basic_table() %>%
#'   split_cols_by(var = "ARM") %>%
#'   analyze_vars(vars = "AVISIT", na.rm = FALSE)
#'
#' build_table(l, df = dta_test)
#'
#' # auto format
#' dt <- data.frame("VAR" = c(0.001, 0.2, 0.0011000, 3, 4))
#' basic_table() %>%
#'   analyze_vars(
#'     vars = "VAR",
#'     .stats = c("n", "mean", "mean_sd", "range"),
#'     .formats = c("mean_sd" = "auto", "range" = "auto")
#'   ) %>%
#'   build_table(dt)
#'
#' @export
#' @order 2
analyze_vars <- function(lyt,
                         vars,
                         var_labels = vars,
                         na_str = default_na_str(),
                         nested = TRUE,
                         ...,
                         na.rm = TRUE, # nolint
                         show_labels = "default",
                         table_names = vars,
                         section_div = NA_character_,
                         .stats = c("n", "mean_sd", "median", "range", "count_fraction"),
                         .formats = NULL,
                         .labels = NULL,
                         .indent_mods = NULL) {
  extra_args <- list(.stats = .stats, na.rm = na.rm, na_str = na_str, ...)
  if (!is.null(.formats)) extra_args[[".formats"]] <- .formats
  if (!is.null(.labels)) extra_args[[".labels"]] <- .labels
  if (!is.null(.indent_mods)) extra_args[[".indent_mods"]] <- .indent_mods

  analyze(
    lyt = lyt,
    vars = vars,
    var_labels = var_labels,
    afun = a_summary,
    na_str = na_str,
    nested = nested,
    extra_args = extra_args,
    inclNAs = TRUE,
    show_labels = show_labels,
    table_names = table_names,
    section_div = section_div
  )
}

#' Kaplan-Meier plot
#'
#' @description `r lifecycle::badge("stable")`
#'
#' From a survival model, a graphic is rendered along with tabulated annotation
#' including the number of patient at risk at given time and the median survival
#' per group.
#'
#' @inheritParams argument_convention
#' @param variables (named `list`)\cr variable names. Details are:
#'   * `tte` (`numeric`)\cr variable indicating time-to-event duration values.
#'   * `is_event` (`logical`)\cr event variable. `TRUE` if event, `FALSE` if time to event is censored.
#'   * `arm` (`factor`)\cr the treatment group variable.
#'   * `strata` (`character` or `NULL`)\cr variable names indicating stratification factors.
#' @param control_surv (`list`)\cr parameters for comparison details, specified by using
#'   the helper function [control_surv_timepoint()]. Some possible parameter options are:
#'   * `conf_level` (`proportion`)\cr confidence level of the interval for survival rate.
#'   * `conf_type` (`string`)\cr `"plain"` (default), `"log"`, `"log-log"` for confidence interval type,
#'     see more in [survival::survfit()]. Note that the option "none" is no longer supported.
#' @param col (`character`)\cr lines colors. Length of a vector should be equal
#'   to number of strata from [survival::survfit()].
#' @param lty (`numeric`)\cr line type. If a vector is given, its length should be equal to the number of strata from
#'   [survival::survfit()].
#' @param lwd (`numeric`)\cr line width. If a vector is given, its length should be equal to the number of strata from
#'   [survival::survfit()].
#' @param censor_show (`flag`)\cr whether to show censored observations.
#' @param pch (`string`)\cr name of symbol or character to use as point symbol to indicate censored cases.
#' @param size (`numeric(1)`)\cr size of censored point symbols.
#' @param max_time (`numeric(1)`)\cr maximum value to show on x-axis. Only data values less than or up to
#'   this threshold value will be plotted (defaults to `NULL`).
#' @param xticks (`numeric` or `NULL`)\cr numeric vector of tick positions or a single number with spacing
#'   between ticks on the x-axis. If `NULL` (default), [labeling::extended()] is used to determine
#'   optimal tick positions on the x-axis.
#' @param xlab (`string`)\cr x-axis label.
#' @param yval (`string`)\cr type of plot, to be plotted on the y-axis. Options are `Survival` (default) and `Failure`
#'   probability.
#' @param ylab (`string`)\cr y-axis label.
#' @param title (`string`)\cr plot title.
#' @param footnotes (`string`)\cr plot footnotes.
#' @param font_size (`numeric(1)`)\cr font size to use for all text.
#' @param ci_ribbon (`flag`)\cr whether the confidence interval should be drawn around the Kaplan-Meier curve.
#' @param annot_at_risk (`flag`)\cr compute and add the annotation table reporting the number of patient at risk
#'   matching the main grid of the Kaplan-Meier curve.
#' @param annot_at_risk_title (`flag`)\cr whether the "Patients at Risk" title should be added above the `annot_at_risk`
#'   table. Has no effect if `annot_at_risk` is `FALSE`. Defaults to `TRUE`.
#' @param annot_surv_med (`flag`)\cr compute and add the annotation table on the Kaplan-Meier curve estimating the
#'   median survival time per group.
#' @param annot_coxph (`flag`)\cr whether to add the annotation table from a [survival::coxph()] model.
#' @param annot_stats (`string` or `NULL`)\cr statistics annotations to add to the plot. Options are
#'   `median` (median survival follow-up time) and `min` (minimum survival follow-up time).
#' @param annot_stats_vlines (`flag`)\cr add vertical lines corresponding to each of the statistics
#'   specified by `annot_stats`. If `annot_stats` is `NULL` no lines will be added.
#' @param control_coxph_pw (`list`)\cr parameters for comparison details, specified using the helper function
#'   [control_coxph()]. Some possible parameter options are:
#'   * `pval_method` (`string`)\cr p-value method for testing hazard ratio = 1.
#'     Default method is `"log-rank"`, can also be set to `"wald"` or `"likelihood"`.
#'   * `ties` (`string`)\cr method for tie handling. Default is `"efron"`,
#'     can also be set to `"breslow"` or `"exact"`. See more in [survival::coxph()]
#'   * `conf_level` (`proportion`)\cr confidence level of the interval for HR.
#' @param ref_group_coxph (`string` or `NULL`)\cr level of arm variable to use as reference group in calculations for
#'   `annot_coxph` table. If `NULL` (default), uses the first level of the arm variable.
#' @param control_annot_surv_med (`list`)\cr parameters to control the position and size of the annotation table added
#'   to the plot when `annot_surv_med = TRUE`, specified using the [control_surv_med_annot()] function. Parameter
#'   options are: `x`, `y`, `w`, `h`, and `fill`. See [control_surv_med_annot()] for details.
#' @param control_annot_coxph (`list`)\cr parameters to control the position and size of the annotation table added
#'   to the plot when `annot_coxph = TRUE`, specified using the [control_coxph_annot()] function. Parameter
#'   options are: `x`, `y`, `w`, `h`, `fill`, and `ref_lbls`. See [control_coxph_annot()] for details.
#' @param legend_pos (`numeric(2)` or `NULL`)\cr vector containing x- and y-coordinates, respectively, for the legend
#'   position relative to the KM plot area. If `NULL` (default), the legend is positioned in the bottom right corner of
#'   the plot, or the middle right of the plot if needed to prevent overlapping.
#' @param rel_height_plot (`proportion`)\cr proportion of total figure height to allocate to the Kaplan-Meier plot.
#'   Relative height of patients at risk table is then `1 - rel_height_plot`. If `annot_at_risk = FALSE` or
#'   `as_list = TRUE`, this parameter is ignored.
#' @param ggtheme (`theme`)\cr a graphical theme as provided by `ggplot2` to format the Kaplan-Meier plot.
#' @param as_list (`flag`)\cr whether the two `ggplot` objects should be returned as a list when `annot_at_risk = TRUE`.
#'   If `TRUE`, a named list with two elements, `plot` and `table`, will be returned. If `FALSE` (default) the patients
#'   at risk table is printed below the plot via [cowplot::plot_grid()].
#' @param draw `r lifecycle::badge("deprecated")` This function no longer generates `grob` objects.
#' @param newpage `r lifecycle::badge("deprecated")` This function no longer generates `grob` objects.
#' @param gp `r lifecycle::badge("deprecated")` This function no longer generates `grob` objects.
#' @param vp `r lifecycle::badge("deprecated")` This function no longer generates `grob` objects.
#' @param name `r lifecycle::badge("deprecated")` This function no longer generates `grob` objects.
#' @param annot_coxph_ref_lbls `r lifecycle::badge("deprecated")` Please use the `ref_lbls` element of
#'   `control_annot_coxph` instead.
#' @param position_coxph `r lifecycle::badge("deprecated")`  Please use the `x` and `y` elements of
#'   `control_annot_coxph` instead.
#' @param position_surv_med `r lifecycle::badge("deprecated")` Please use the `x` and `y` elements of
#'   `control_annot_surv_med` instead.
#' @param width_annots `r lifecycle::badge("deprecated")` Please use the `w` element of `control_annot_surv_med`
#'   (for `surv_med`) and `control_annot_coxph` (for `coxph`)."
#'
#' @return A `ggplot` Kaplan-Meier plot and (optionally) summary table.
#'
#' @examples
#' library(dplyr)
#' library(nestcolor)
#'
#' df <- tern_ex_adtte %>%
#'   filter(PARAMCD == "OS") %>%
#'   mutate(is_event = CNSR == 0)
#' variables <- list(tte = "AVAL", is_event = "is_event", arm = "ARMCD")
#'
#' # Basic examples
#' g_km(df = df, variables = variables)
#' g_km(df = df, variables = variables, yval = "Failure")
#'
#' # Examples with customization parameters applied
#' g_km(
#'   df = df,
#'   variables = variables,
#'   control_surv = control_surv_timepoint(conf_level = 0.9),
#'   col = c("grey25", "grey50", "grey75"),
#'   annot_at_risk_title = FALSE,
#'   lty = 1:3,
#'   font_size = 8
#' )
#' g_km(
#'   df = df,
#'   variables = variables,
#'   annot_stats = c("min", "median"),
#'   annot_stats_vlines = TRUE,
#'   max_time = 3000,
#'   ggtheme = ggplot2::theme_minimal()
#' )
#'
#' # Example with pairwise Cox-PH analysis annotation table, adjusted annotation tables
#' g_km(
#'   df = df, variables = variables,
#'   annot_coxph = TRUE,
#'   control_coxph = control_coxph(pval_method = "wald", ties = "exact", conf_level = 0.99),
#'   control_annot_coxph = control_coxph_annot(x = 0.26, w = 0.35),
#'   control_annot_surv_med = control_surv_med_annot(x = 0.8, y = 0.9, w = 0.35)
#' )
#'
#' @aliases kaplan_meier
#' @export
g_km <- function(df,
                 variables,
                 control_surv = control_surv_timepoint(),
                 col = NULL,
                 lty = NULL,
                 lwd = 0.5,
                 censor_show = TRUE,
                 pch = 3,
                 size = 2,
                 max_time = NULL,
                 xticks = NULL,
                 xlab = "Days",
                 yval = c("Survival", "Failure"),
                 ylab = paste(yval, "Probability"),
                 ylim = NULL,
                 title = NULL,
                 footnotes = NULL,
                 font_size = 10,
                 ci_ribbon = FALSE,
                 annot_at_risk = TRUE,
                 annot_at_risk_title = TRUE,
                 annot_surv_med = TRUE,
                 annot_coxph = FALSE,
                 annot_stats = NULL,
                 annot_stats_vlines = FALSE,
                 control_coxph_pw = control_coxph(),
                 ref_group_coxph = NULL,
                 control_annot_surv_med = control_surv_med_annot(),
                 control_annot_coxph = control_coxph_annot(),
                 legend_pos = NULL,
                 rel_height_plot = 0.75,
                 ggtheme = NULL,
                 as_list = FALSE,
                 draw = lifecycle::deprecated(),
                 newpage = lifecycle::deprecated(),
                 gp = lifecycle::deprecated(),
                 vp = lifecycle::deprecated(),
                 name = lifecycle::deprecated(),
                 annot_coxph_ref_lbls = lifecycle::deprecated(),
                 position_coxph = lifecycle::deprecated(),
                 position_surv_med = lifecycle::deprecated(),
                 width_annots = lifecycle::deprecated()) {
  # Deprecated argument warnings
  if (lifecycle::is_present(draw)) {
    lifecycle::deprecate_warn(
      "0.9.4", "g_km(draw)",
      details = "This argument is no longer used since the plot is now generated as a `ggplot2` object."
    )
  }
  if (lifecycle::is_present(newpage)) {
    lifecycle::deprecate_warn(
      "0.9.4", "g_km(newpage)",
      details = "This argument is no longer used since the plot is now generated as a `ggplot2` object."
    )
  }
  if (lifecycle::is_present(gp)) {
    lifecycle::deprecate_warn(
      "0.9.4", "g_km(gp)",
      details = "This argument is no longer used since the plot is now generated as a `ggplot2` object."
    )
  }
  if (lifecycle::is_present(vp)) {
    lifecycle::deprecate_warn(
      "0.9.4", "g_km(vp)",
      details = "This argument is no longer used since the plot is now generated as a `ggplot2` object."
    )
  }
  if (lifecycle::is_present(name)) {
    lifecycle::deprecate_warn(
      "0.9.4", "g_km(name)",
      details = "This argument is no longer used since the plot is now generated as a `ggplot2` object."
    )
  }
  if (lifecycle::is_present(annot_coxph_ref_lbls)) {
    lifecycle::deprecate_warn(
      "0.9.4", "g_km(annot_coxph_ref_lbls)",
      details = "Please specify this setting using the 'ref_lbls' element of control_annot_coxph."
    )
    control_annot_coxph[["ref_lbls"]] <- annot_coxph_ref_lbls
  }
  if (lifecycle::is_present(position_coxph)) {
    lifecycle::deprecate_warn(
      "0.9.4", "g_km(position_coxph)",
      details = "Please specify this setting using the 'x' and 'y' elements of control_annot_coxph."
    )
    control_annot_coxph[["x"]] <- position_coxph[1]
    control_annot_coxph[["y"]] <- position_coxph[2]
  }
  if (lifecycle::is_present(position_surv_med)) {
    lifecycle::deprecate_warn(
      "0.9.4", "g_km(position_surv_med)",
      details = "Please specify this setting using the 'x' and 'y' elements of control_annot_surv_med."
    )
    control_annot_surv_med[["x"]] <- position_surv_med[1]
    control_annot_surv_med[["y"]] <- position_surv_med[2]
  }
  if (lifecycle::is_present(width_annots)) {
    lifecycle::deprecate_warn(
      "0.9.4", "g_km(width_annots)",
      details = paste(
        "Please specify widths of annotation tables relative to the plot area using the 'w' element of",
        "control_annot_surv_med (for surv_med) and control_annot_coxph (for coxph)."
      )
    )
    control_annot_surv_med[["w"]] <- as.numeric(width_annots[["surv_med"]])
    control_annot_coxph[["w"]] <- as.numeric(width_annots[["coxph"]])
  }

  checkmate::assert_list(variables)
  checkmate::assert_subset(c("tte", "arm", "is_event"), names(variables))
  checkmate::assert_logical(censor_show, len = 1)
  checkmate::assert_numeric(size, len = 1)
  checkmate::assert_numeric(max_time, len = 1, null.ok = TRUE)
  checkmate::assert_numeric(xticks, null.ok = TRUE)
  checkmate::assert_character(xlab, len = 1, null.ok = TRUE)
  checkmate::assert_character(yval)
  checkmate::assert_character(ylab, null.ok = TRUE)
  checkmate::assert_numeric(ylim, finite = TRUE, any.missing = FALSE, len = 2, sorted = TRUE, null.ok = TRUE)
  checkmate::assert_character(title, len = 1, null.ok = TRUE)
  checkmate::assert_character(footnotes, len = 1, null.ok = TRUE)
  checkmate::assert_numeric(font_size, len = 1)
  checkmate::assert_logical(ci_ribbon, len = 1)
  checkmate::assert_logical(annot_at_risk, len = 1)
  checkmate::assert_logical(annot_at_risk_title, len = 1)
  checkmate::assert_logical(annot_surv_med, len = 1)
  checkmate::assert_logical(annot_coxph, len = 1)
  checkmate::assert_subset(annot_stats, c("median", "min"))
  checkmate::assert_logical(annot_stats_vlines)
  checkmate::assert_list(control_coxph_pw)
  checkmate::assert_character(ref_group_coxph, len = 1, null.ok = TRUE)
  checkmate::assert_list(control_annot_surv_med)
  checkmate::assert_list(control_annot_coxph)
  checkmate::assert_numeric(legend_pos, finite = TRUE, any.missing = FALSE, len = 2, null.ok = TRUE)
  assert_proportion_value(rel_height_plot)
  checkmate::assert_logical(as_list)

  tte <- variables$tte
  is_event <- variables$is_event
  arm <- variables$arm
  assert_valid_factor(df[[arm]])
  armval <- as.character(unique(df[[arm]]))
  assert_df_with_variables(df, list(tte = tte, is_event = is_event, arm = arm))
  checkmate::assert_logical(df[[is_event]], min.len = 1)
  checkmate::assert_numeric(df[[tte]], min.len = 1)
  checkmate::assert_vector(col, len = length(armval), null.ok = TRUE)
  checkmate::assert_vector(lty, null.ok = TRUE)
  checkmate::assert_numeric(lwd, len = 1, null.ok = TRUE)

  if (annot_coxph && length(armval) < 2) {
    stop(paste(
      "When `annot_coxph` = TRUE, `df` must contain at least 2 levels of `variables$arm`",
      "in order to calculate the hazard ratio."
    ))
  }

  # process model
  yval <- match.arg(yval)
  formula <- stats::as.formula(paste0("survival::Surv(", tte, ", ", is_event, ") ~ ", arm))
  fit_km <- survival::survfit(
    formula = formula,
    data = df,
    conf.int = control_surv$conf_level,
    conf.type = control_surv$conf_type
  )
  data <- h_data_plot(fit_km, armval = armval, max_time = max_time)

  # calculate x-ticks
  xticks <- h_xticks(data = data, xticks = xticks, max_time = max_time)

  # change estimates of survival to estimates of failure (1 - survival)
  if (yval == "Failure") {
    data[c("estimate", "conf.low", "conf.high", "censor")] <- list(
      1 - data$estimate, 1 - data$conf.low, 1 - data$conf.high, 1 - data$censor
    )
  }

  # derive y-axis limits
  if (is.null(ylim)) {
    if (!is.null(max_time)) {
      y_lwr <- min(data[data$time < max_time, ][["estimate"]])
      y_upr <- max(data[data$time < max_time, ][["estimate"]])
    } else {
      y_lwr <- min(data[["estimate"]])
      y_upr <- max(data[["estimate"]])
    }
    ylim <- c(y_lwr, y_upr)
  }

  # initialize ggplot
  gg_plt <- ggplot(
    data = data,
    mapping = aes(
      x = .data[["time"]],
      y = .data[["estimate"]],
      ymin = .data[["conf.low"]],
      ymax = .data[["conf.high"]],
      color = .data[["strata"]],
      fill = .data[["strata"]]
    )
  ) +
    theme_bw(base_size = font_size) +
    scale_y_continuous(limits = ylim, expand = c(0.025, 0)) +
    labs(title = title, x = xlab, y = ylab, caption = footnotes) +
    theme(
      axis.text = element_text(size = font_size),
      axis.title = element_text(size = font_size),
      legend.title = element_blank(),
      legend.text = element_text(size = font_size),
      legend.box.background = element_rect(fill = "white", linewidth = 0.5),
      legend.background = element_blank(),
      legend.position = "inside",
      legend.spacing.y = unit(-0.02, "npc"),
      panel.grid.major = element_blank(),
      panel.grid.minor = element_blank()
    )

  # derive x-axis limits
  if (!is.null(max_time) && !is.null(xticks)) {
    gg_plt <- gg_plt + scale_x_continuous(
      breaks = xticks, limits = c(min(0, xticks), max(c(xticks, max_time))), expand = c(0.025, 0)
    )
  } else if (!is.null(xticks)) {
    if (max(data$time) <= max(xticks)) {
      gg_plt <- gg_plt + scale_x_continuous(
        breaks = xticks, limits = c(min(0, min(xticks)), max(xticks)), expand = c(0.025, 0)
      )
    } else {
      gg_plt <- gg_plt + scale_x_continuous(breaks = xticks, expand = c(0.025, 0))
    }
  } else if (!is.null(max_time)) {
    gg_plt <- gg_plt + scale_x_continuous(limits = c(0, max_time), expand = c(0.025, 0))
  }

  # set legend position
  if (!is.null(legend_pos)) {
    gg_plt <- gg_plt + theme(legend.position.inside = legend_pos)
  } else {
    max_time2 <- sort(
      data$time,
      partial = nrow(data) - length(armval) - 1
    )[nrow(data) - length(armval) - 1]

    y_rng <- ylim[2] - ylim[1]

    if (yval == "Survival" && all(data$estimate[data$time == max_time2] > ylim[1] + 0.09 * y_rng) &&
      all(data$estimate[data$time == max_time2] < ylim[1] + 0.5 * y_rng)) { # nolint
      gg_plt <- gg_plt +
        theme(
          legend.position.inside = c(1, 0.5),
          legend.justification = c(1.1, 0.6)
        )
    } else {
      gg_plt <- gg_plt +
        theme(
          legend.position.inside = c(1, 0),
          legend.justification = c(1.1, -0.4)
        )
    }
  }

  # add lines
  gg_plt <- if (is.null(lty)) {
    gg_plt + geom_step(linewidth = lwd, na.rm = TRUE)
  } else if (length(lty) == 1) {
    gg_plt + geom_step(linewidth = lwd, lty = lty, na.rm = TRUE)
  } else {
    gg_plt +
      geom_step(aes(lty = .data[["strata"]]), linewidth = lwd, na.rm = TRUE) +
      scale_linetype_manual(values = lty)
  }

  # add censor marks
  if (censor_show) {
    gg_plt <- gg_plt + geom_point(
      data = data[data$n.censor != 0, ],
      aes(x = .data[["time"]], y = .data[["censor"]], shape = "Censored"),
      size = size,
      na.rm = TRUE
    ) +
      scale_shape_manual(name = NULL, values = pch) +
      guides(fill = guide_legend(override.aes = list(shape = NA)))
  }

  # add ci ribbon
  if (ci_ribbon) gg_plt <- gg_plt + geom_ribbon(alpha = 0.3, lty = 0, na.rm = TRUE)

  # control aesthetics
  if (!is.null(col)) {
    gg_plt <- gg_plt +
      scale_color_manual(values = col) +
      scale_fill_manual(values = col)
  }
  if (!is.null(ggtheme)) gg_plt <- gg_plt + ggtheme

  # annotate with stats (text/vlines)
  if (!is.null(annot_stats)) {
    if ("median" %in% annot_stats) {
      fit_km_all <- survival::survfit(
        formula = stats::as.formula(paste0("survival::Surv(", tte, ", ", is_event, ") ~ ", 1)),
        data = df,
        conf.int = control_surv$conf_level,
        conf.type = control_surv$conf_type
      )
      gg_plt <- gg_plt +
        annotate(
          "text",
          size = font_size / .pt, col = 1, lineheight = 0.95,
          x = stats::median(fit_km_all) + 0.07 * max(data$time),
          y = ifelse(yval == "Survival", 0.65, 0.35),
          label = paste("Median F/U:\n", round(stats::median(fit_km_all), 1), tolower(df$AVALU[1]))
        )
      if (annot_stats_vlines) {
        gg_plt <- gg_plt +
          annotate(
            "segment",
            x = stats::median(fit_km_all), xend = stats::median(fit_km_all), y = -Inf, yend = Inf,
            linetype = 2, col = "darkgray"
          )
      }
    }
    if ("min" %in% annot_stats) {
      min_fu <- min(df[[tte]])
      gg_plt <- gg_plt +
        annotate(
          "text",
          size = font_size / .pt, col = 1, lineheight = 0.95,
          x = min_fu + max(data$time) * 0.07,
          y = ifelse(yval == "Survival", 0.96, 0.05),
          label = paste("Min. F/U:\n", round(min_fu, 1), tolower(df$AVALU[1]))
        )
      if (annot_stats_vlines) {
        gg_plt <- gg_plt +
          annotate(
            "segment",
            linetype = 2, col = "darkgray",
            x = min_fu, xend = min_fu, y = Inf, yend = -Inf
          )
      }
    }
    gg_plt <- gg_plt + guides(fill = guide_legend(override.aes = list(shape = NA, label = "")))
  }

  # add at risk annotation table
  if (annot_at_risk) {
    annot_tbl <- summary(fit_km, times = xticks, extend = TRUE)
    annot_tbl <- if (is.null(fit_km$strata)) {
      data.frame(
        n.risk = annot_tbl$n.risk,
        time = annot_tbl$time,
        strata = armval
      )
    } else {
      strata_lst <- strsplit(sub("=", "equals", levels(annot_tbl$strata)), "equals")
      levels(annot_tbl$strata) <- matrix(unlist(strata_lst), ncol = 2, byrow = TRUE)[, 2]
      data.frame(
        n.risk = annot_tbl$n.risk,
        time = annot_tbl$time,
        strata = annot_tbl$strata
      )
    }

    at_risk_tbl <- as.data.frame(tidyr::pivot_wider(annot_tbl, names_from = "time", values_from = "n.risk")[, -1])
    at_risk_tbl[is.na(at_risk_tbl)] <- 0
    rownames(at_risk_tbl) <- levels(annot_tbl$strata)

    gg_at_risk <- df2gg(
      at_risk_tbl,
      font_size = font_size, col_labels = FALSE, hline = FALSE,
      colwidths = rep(1, ncol(at_risk_tbl))
    ) +
      labs(title = if (annot_at_risk_title) "Patients at Risk:" else NULL, x = xlab) +
      theme_bw(base_size = font_size) +
      theme(
        plot.title = element_text(size = font_size, vjust = 3, face = "bold"),
        panel.border = element_blank(),
        panel.grid = element_blank(),
        axis.title.y = element_blank(),
        axis.ticks.y = element_blank(),
        axis.text.y = element_text(size = font_size, face = "italic", hjust = 1),
        axis.text.x = element_text(size = font_size),
        axis.line.x = element_line()
      ) +
      coord_cartesian(clip = "off", ylim = c(0.5, nrow(at_risk_tbl)))
    gg_at_risk <- suppressMessages(
      gg_at_risk +
        scale_x_continuous(expand = c(0.025, 0), breaks = seq_along(at_risk_tbl) - 0.5, labels = xticks) +
        scale_y_continuous(labels = rev(levels(annot_tbl$strata)), breaks = seq_len(nrow(at_risk_tbl)))
    )

    if (!as_list) {
      gg_plt <- cowplot::plot_grid(
        gg_plt,
        gg_at_risk,
        align = "v",
        axis = "tblr",
        ncol = 1,
        rel_heights = c(rel_height_plot, 1 - rel_height_plot)
      )
    }
  }

  # add median survival time annotation table
  if (annot_surv_med) {
    surv_med_tbl <- h_tbl_median_surv(fit_km = fit_km, armval = armval)
    bg_fill <- if (isTRUE(control_annot_surv_med[["fill"]])) "#00000020" else control_annot_surv_med[["fill"]]

    gg_surv_med <- df2gg(surv_med_tbl, font_size = font_size, colwidths = c(1, 1, 2), bg_fill = bg_fill) +
      theme(
        axis.text.y = element_text(size = font_size, face = "italic", hjust = 1),
        plot.margin = margin(0, 2, 0, 5)
      ) +
      coord_cartesian(clip = "off", ylim = c(0.5, nrow(surv_med_tbl) + 1.5))
    gg_surv_med <- suppressMessages(
      gg_surv_med +
        scale_x_continuous(expand = c(0.025, 0)) +
        scale_y_continuous(labels = rev(rownames(surv_med_tbl)), breaks = seq_len(nrow(surv_med_tbl)))
    )

    gg_plt <- cowplot::ggdraw(gg_plt) +
      cowplot::draw_plot(
        gg_surv_med,
        control_annot_surv_med[["x"]],
        control_annot_surv_med[["y"]],
        width = control_annot_surv_med[["w"]],
        height = control_annot_surv_med[["h"]],
        vjust = 0.5,
        hjust = 0.5
      )
  }

  # add coxph annotation table
  if (annot_coxph) {
    coxph_tbl <- h_tbl_coxph_pairwise(
      df = df,
      variables = variables,
      ref_group_coxph = ref_group_coxph,
      control_coxph_pw = control_coxph_pw,
      annot_coxph_ref_lbls = control_annot_coxph[["ref_lbls"]]
    )
    bg_fill <- if (isTRUE(control_annot_coxph[["fill"]])) "#00000020" else control_annot_coxph[["fill"]]

    gg_coxph <- df2gg(coxph_tbl, font_size = font_size, colwidths = c(1.1, 1, 3), bg_fill = bg_fill) +
      theme(
        axis.text.y = element_text(size = font_size, face = "italic", hjust = 1),
        plot.margin = margin(0, 2, 0, 5)
      ) +
      coord_cartesian(clip = "off", ylim = c(0.5, nrow(coxph_tbl) + 1.5))
    gg_coxph <- suppressMessages(
      gg_coxph +
        scale_x_continuous(expand = c(0.025, 0)) +
        scale_y_continuous(labels = rev(rownames(coxph_tbl)), breaks = seq_len(nrow(coxph_tbl)))
    )

    gg_plt <- cowplot::ggdraw(gg_plt) +
      cowplot::draw_plot(
        gg_coxph,
        control_annot_coxph[["x"]],
        control_annot_coxph[["y"]],
        width = control_annot_coxph[["w"]],
        height = control_annot_coxph[["h"]],
        vjust = 0.5,
        hjust = 0.5
      )
  }

  if (as_list) {
    list(plot = gg_plt, table = gg_at_risk)
  } else {
    gg_plt
  }
}

#' Summarize analysis of covariance (ANCOVA) results
#'
#' @description `r lifecycle::badge("stable")`
#'
#' The analyze function [summarize_ancova()] creates a layout element to summarize ANCOVA results.
#'
#' This function can be used to analyze multiple endpoints and/or multiple timepoints within the response variable(s)
#' specified as `vars`.
#'
#' Additional variables for the analysis, namely an arm (grouping) variable and covariate variables, can be defined
#' via the `variables` argument. See below for more details on how to specify `variables`. An interaction term can
#' be implemented in the model if needed. The interaction variable that should interact with the arm variable is
#' specified via the `interaction_term` parameter, and the specific value of `interaction_term` for which to extract
#' the ANCOVA results via the `interaction_y` parameter.
#'
#' @inheritParams h_ancova
#' @inheritParams argument_convention
#' @param interaction_y (`string` or `flag`)\cr a selected item inside of the `interaction_item` variable which will be
#'   used to select the specific ANCOVA results. if the interaction is not needed, the default option is `FALSE`.
#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("summarize_ancova")`
#'   to see available statistics for this function.
#'
#' @name summarize_ancova
#' @order 1
NULL

#' Helper function to return results of a linear model
#'
#' @description `r lifecycle::badge("stable")`
#'
#' @inheritParams argument_convention
#' @param .df_row (`data.frame`)\cr data set that includes all the variables that are called in `.var` and `variables`.
#' @param variables (named `list` of `string`)\cr list of additional analysis variables, with expected elements:
#'   * `arm` (`string`)\cr group variable, for which the covariate adjusted means of multiple groups will be
#'     summarized. Specifically, the first level of `arm` variable is taken as the reference group.
#'   * `covariates` (`character`)\cr a vector that can contain single variable names (such as `"X1"`), and/or
#'     interaction terms indicated by `"X1 * X2"`.
#' @param interaction_item (`string` or `NULL`)\cr name of the variable that should have interactions
#'   with arm. if the interaction is not needed, the default option is `NULL`.
#'
#' @return The summary of a linear model.
#'
#' @examples
#' h_ancova(
#'   .var = "Sepal.Length",
#'   .df_row = iris,
#'   variables = list(arm = "Species", covariates = c("Petal.Length * Petal.Width", "Sepal.Width"))
#' )
#'
#' @export
h_ancova <- function(.var,
                     .df_row,
                     variables,
                     interaction_item = NULL) {
  checkmate::assert_string(.var)
  checkmate::assert_list(variables)
  checkmate::assert_subset(names(variables), c("arm", "covariates"))
  assert_df_with_variables(.df_row, list(rsp = .var))

  arm <- variables$arm
  covariates <- variables$covariates
  if (!is.null(covariates) && length(covariates) > 0) {
    # Get all covariate variable names in the model.
    var_list <- get_covariates(covariates)
    assert_df_with_variables(.df_row, var_list)
  }

  covariates_part <- paste(covariates, collapse = " + ")
  if (covariates_part != "") {
    formula <- stats::as.formula(paste0(.var, " ~ ", covariates_part, " + ", arm))
  } else {
    formula <- stats::as.formula(paste0(.var, " ~ ", arm))
  }

  if (is.null(interaction_item)) {
    specs <- arm
  } else {
    specs <- c(arm, interaction_item)
  }

  lm_fit <- stats::lm(
    formula = formula,
    data = .df_row
  )
  emmeans_fit <- emmeans::emmeans(
    lm_fit,
    # Specify here the group variable over which EMM are desired.
    specs = specs,
    # Pass the data again so that the factor levels of the arm variable can be inferred.
    data = .df_row
  )

  emmeans_fit
}

#' @describeIn summarize_ancova Statistics function that produces a named list of results
#'   of the investigated linear model.
#'
#' @return
#' * `s_ancova()` returns a named list of 5 statistics:
#'   * `n`: Count of complete sample size for the group.
#'   * `lsmean`: Estimated marginal means in the group.
#'   * `lsmean_diff`: Difference in estimated marginal means in comparison to the reference group.
#'     If working with the reference group, this will be empty.
#'   * `lsmean_diff_ci`: Confidence level for difference in estimated marginal means in comparison
#'     to the reference group.
#'   * `pval`: p-value (not adjusted for multiple comparisons).
#'
#' @keywords internal
s_ancova <- function(df,
                     .var,
                     .df_row,
                     variables,
                     .ref_group,
                     .in_ref_col,
                     conf_level,
                     interaction_y = FALSE,
                     interaction_item = NULL) {
  emmeans_fit <- h_ancova(.var = .var, variables = variables, .df_row = .df_row, interaction_item = interaction_item)

  sum_fit <- summary(
    emmeans_fit,
    level = conf_level
  )

  arm <- variables$arm

  sum_level <- as.character(unique(df[[arm]]))

  # Ensure that there is only one element in sum_level.
  checkmate::assert_scalar(sum_level)

  sum_fit_level <- sum_fit[sum_fit[[arm]] == sum_level, ]

  # Get the index of the ref arm
  if (interaction_y != FALSE) {
    y <- unlist(df[(df[[interaction_item]] == interaction_y), .var])
    # convert characters selected in interaction_y into the numeric order
    interaction_y <- which(sum_fit_level[[interaction_item]] == interaction_y)
    sum_fit_level <- sum_fit_level[interaction_y, ]
    # if interaction is called, reset the index
    ref_key <- seq(sum_fit[[arm]][unique(.ref_group[[arm]])])
    ref_key <- tail(ref_key, n = 1)
    ref_key <- (interaction_y - 1) * length(unique(.df_row[[arm]])) + ref_key
  } else {
    y <- df[[.var]]
    # Get the index of the ref arm when interaction is not called
    ref_key <- seq(sum_fit[[arm]][unique(.ref_group[[arm]])])
    ref_key <- tail(ref_key, n = 1)
  }

  if (.in_ref_col) {
    list(
      n = length(y[!is.na(y)]),
      lsmean = formatters::with_label(sum_fit_level$emmean, "Adjusted Mean"),
      lsmean_diff = formatters::with_label(character(), "Difference in Adjusted Means"),
      lsmean_diff_ci = formatters::with_label(character(), f_conf_level(conf_level)),
      pval = formatters::with_label(character(), "p-value")
    )
  } else {
    # Estimate the differences between the marginal means.
    emmeans_contrasts <- emmeans::contrast(
      emmeans_fit,
      # Compare all arms versus the control arm.
      method = "trt.vs.ctrl",
      # Take the arm factor from .ref_group as the control arm.
      ref = ref_key,
      level = conf_level
    )
    sum_contrasts <- summary(
      emmeans_contrasts,
      # Derive confidence intervals, t-tests and p-values.
      infer = TRUE,
      # Do not adjust the p-values for multiplicity.
      adjust = "none"
    )

    contrast_lvls <- gsub(
      "^\\(|\\)$", "", gsub(paste0(" - \\(*", .ref_group[[arm]][1], ".*"), "", sum_contrasts$contrast)
    )
    if (!is.null(interaction_item)) {
      sum_contrasts_level <- sum_contrasts[grepl(sum_level, contrast_lvls, fixed = TRUE), ]
    } else {
      sum_contrasts_level <- sum_contrasts[sum_level == contrast_lvls, ]
    }
    if (interaction_y != FALSE) {
      sum_contrasts_level <- sum_contrasts_level[interaction_y, ]
    }

    list(
      n = length(y[!is.na(y)]),
      lsmean = formatters::with_label(sum_fit_level$emmean, "Adjusted Mean"),
      lsmean_diff = formatters::with_label(sum_contrasts_level$estimate, "Difference in Adjusted Means"),
      lsmean_diff_ci = formatters::with_label(
        c(sum_contrasts_level$lower.CL, sum_contrasts_level$upper.CL),
        f_conf_level(conf_level)
      ),
      pval = formatters::with_label(sum_contrasts_level$p.value, "p-value")
    )
  }
}

#' @describeIn summarize_ancova Formatted analysis function which is used as `afun` in `summarize_ancova()`.
#'
#' @return
#' * `a_ancova()` returns the corresponding list with formatted [rtables::CellValue()].
#'
#' @keywords internal
a_ancova <- make_afun(
  s_ancova,
  .indent_mods = c("n" = 0L, "lsmean" = 0L, "lsmean_diff" = 0L, "lsmean_diff_ci" = 1L, "pval" = 1L),
  .formats = c(
    "n" = "xx",
    "lsmean" = "xx.xx",
    "lsmean_diff" = "xx.xx",
    "lsmean_diff_ci" = "(xx.xx, xx.xx)",
    "pval" = "x.xxxx | (<0.0001)"
  ),
  .null_ref_cells = FALSE
)

#' @describeIn summarize_ancova Layout-creating function which can take statistics function arguments
#'   and additional format arguments. This function is a wrapper for [rtables::analyze()].
#'
#' @return
#' * `summarize_ancova()` returns a layout object suitable for passing to further layouting functions,
#'   or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
#'   the statistics from `s_ancova()` to the table layout.
#'
#' @examples
#' basic_table() %>%
#'   split_cols_by("Species", ref_group = "setosa") %>%
#'   add_colcounts() %>%
#'   summarize_ancova(
#'     vars = "Petal.Length",
#'     variables = list(arm = "Species", covariates = NULL),
#'     table_names = "unadj",
#'     conf_level = 0.95, var_labels = "Unadjusted comparison",
#'     .labels = c(lsmean = "Mean", lsmean_diff = "Difference in Means")
#'   ) %>%
#'   summarize_ancova(
#'     vars = "Petal.Length",
#'     variables = list(arm = "Species", covariates = c("Sepal.Length", "Sepal.Width")),
#'     table_names = "adj",
#'     conf_level = 0.95, var_labels = "Adjusted comparison (covariates: Sepal.Length and Sepal.Width)"
#'   ) %>%
#'   build_table(iris)
#'
#' @export
#' @order 2
summarize_ancova <- function(lyt,
                             vars,
                             variables,
                             conf_level,
                             interaction_y = FALSE,
                             interaction_item = NULL,
                             var_labels,
                             na_str = default_na_str(),
                             nested = TRUE,
                             ...,
                             show_labels = "visible",
                             table_names = vars,
                             .stats = NULL,
                             .formats = NULL,
                             .labels = NULL,
                             .indent_mods = NULL) {
  extra_args <- list(
    variables = variables, conf_level = conf_level, interaction_y = interaction_y,
    interaction_item = interaction_item, ...
  )

  afun <- make_afun(
    a_ancova,
    interaction_y = interaction_y,
    interaction_item = interaction_item,
    .stats = .stats,
    .formats = .formats,
    .labels = .labels,
    .indent_mods = .indent_mods
  )

  analyze(
    lyt,
    vars,
    var_labels = var_labels,
    show_labels = show_labels,
    table_names = table_names,
    afun = afun,
    na_str = na_str,
    nested = nested,
    extra_args = extra_args
  )
}

#' Helper functions for accessing information from `rtables`
#'
#' @description `r lifecycle::badge("stable")`
#'
#' These are a couple of functions that help with accessing the data in `rtables` objects.
#' Currently these work for occurrence tables, which are defined as having a count as the first
#' element and a fraction as the second element in each cell.
#'
#' @seealso [prune_occurrences] for usage of these functions.
#'
#' @name rtables_access
NULL

#' @describeIn rtables_access Helper function to extract the first values from each content
#'   cell and from specified columns in a `TableRow`. Defaults to all columns.
#'
#' @param table_row (`TableRow`)\cr an analysis row in a occurrence table.
#' @param col_names (`character`)\cr the names of the columns to extract from.
#' @param col_indices (`integer`)\cr the indices of the columns to extract from. If `col_names` are provided,
#'   then these are inferred from the names of `table_row`. Note that this currently only works well with a single
#'   column split.
#'
#' @return
#' * `h_row_first_values()` returns a `vector` of numeric values.
#'
#' @examples
#' tbl <- basic_table() %>%
#'   split_cols_by("ARM") %>%
#'   split_rows_by("RACE") %>%
#'   analyze("AGE", function(x) {
#'     list(
#'       "mean (sd)" = rcell(c(mean(x), sd(x)), format = "xx.x (xx.x)"),
#'       "n" = length(x),
#'       "frac" = rcell(c(0.1, 0.1), format = "xx (xx)")
#'     )
#'   }) %>%
#'   build_table(tern_ex_adsl) %>%
#'   prune_table()
#' tree_row_elem <- collect_leaves(tbl[2, ])[[1]]
#' result <- max(h_row_first_values(tree_row_elem))
#' result
#'
#' @export
h_row_first_values <- function(table_row,
                               col_names = NULL,
                               col_indices = NULL) {
  col_indices <- check_names_indices(table_row, col_names, col_indices)
  checkmate::assert_integerish(col_indices)
  checkmate::assert_subset(col_indices, seq_len(ncol(table_row)))

  # Main values are extracted
  row_vals <- row_values(table_row)[col_indices]

  # Main return
  vapply(row_vals, function(rv) {
    if (is.null(rv)) {
      NA_real_
    } else {
      rv[1L]
    }
  }, FUN.VALUE = numeric(1))
}

#' @describeIn rtables_access Helper function that extracts row values and checks if they are
#'   convertible to integers (`integerish` values).
#'
#' @return
#' * `h_row_counts()` returns a `vector` of numeric values.
#'
#' @examples
#' # Row counts (integer values)
#' # h_row_counts(tree_row_elem) # Fails because there are no integers
#' # Using values with integers
#' tree_row_elem <- collect_leaves(tbl[3, ])[[1]]
#' result <- h_row_counts(tree_row_elem)
#' # result
#'
#' @export
h_row_counts <- function(table_row,
                         col_names = NULL,
                         col_indices = NULL) {
  counts <- h_row_first_values(table_row, col_names, col_indices)
  checkmate::assert_integerish(counts)
  counts
}

#' @describeIn rtables_access Helper function to extract fractions from specified columns in a `TableRow`.
#'   More specifically it extracts the second values from each content cell and checks it is a fraction.
#'
#' @return
#' * `h_row_fractions()` returns a `vector` of proportions.
#'
#' @examples
#' # Row fractions
#' tree_row_elem <- collect_leaves(tbl[4, ])[[1]]
#' h_row_fractions(tree_row_elem)
#'
#' @export
h_row_fractions <- function(table_row,
                            col_names = NULL,
                            col_indices = NULL) {
  col_indices <- check_names_indices(table_row, col_names, col_indices)
  row_vals <- row_values(table_row)[col_indices]
  fractions <- sapply(row_vals, "[", 2L)
  checkmate::assert_numeric(fractions, lower = 0, upper = 1)
  fractions
}

#' @describeIn rtables_access Helper function to extract column counts from specified columns in a table.
#'
#' @param table (`VTableNodeInfo`)\cr an occurrence table or row.
#'
#' @return
#' * `h_col_counts()` returns a `vector` of column counts.
#'
#' @export
h_col_counts <- function(table,
                         col_names = NULL,
                         col_indices = NULL) {
  col_indices <- check_names_indices(table, col_names, col_indices)
  counts <- col_counts(table)[col_indices]
  stats::setNames(counts, col_names)
}

#' @describeIn rtables_access Helper function to get first row of content table of current table.
#'
#' @return
#' * `h_content_first_row()` returns a row from an `rtables` table.
#'
#' @export
h_content_first_row <- function(table) {
  ct <- content_table(table)
  tree_children(ct)[[1]]
}

#' @describeIn rtables_access Helper function which says whether current table is a leaf in the tree.
#'
#' @return
#' * `is_leaf_table()` returns a `logical` value indicating whether current table is a leaf.
#'
#' @keywords internal
is_leaf_table <- function(table) {
  children <- tree_children(table)
  child_classes <- unique(sapply(children, class))
  identical(child_classes, "ElementaryTable")
}

#' @describeIn rtables_access Internal helper function that tests standard inputs for column indices.
#'
#' @return
#' * `check_names_indices` returns column indices.
#'
#' @keywords internal
check_names_indices <- function(table_row,
                                col_names = NULL,
                                col_indices = NULL) {
  if (!is.null(col_names)) {
    if (!is.null(col_indices)) {
      stop(
        "Inserted both col_names and col_indices when selecting row values. ",
        "Please choose one."
      )
    }
    col_indices <- h_col_indices(table_row, col_names)
  }
  if (is.null(col_indices)) {
    ll <- ifelse(is.null(ncol(table_row)), length(table_row), ncol(table_row))
    col_indices <- seq_len(ll)
  }

  return(col_indices)
}

#' Count patients by most extreme post-baseline toxicity grade per direction of abnormality
#'
#' @description `r lifecycle::badge("stable")`
#'
#' The analyze function [count_abnormal_by_worst_grade()] creates a layout element to count patients by highest (worst)
#' analysis toxicity grade post-baseline for each direction, categorized by parameter value.
#'
#' This function analyzes primary analysis variable `var` which indicates toxicity grades. Additional
#' analysis variables that can be supplied as a list via the `variables` parameter are `id` (defaults to
#' `USUBJID`), a variable to indicate unique subject identifiers, `param` (defaults to `PARAM`), a variable
#' to indicate parameter values, and `grade_dir` (defaults to `GRADE_DIR`), a variable to indicate directions
#' (e.g. High or Low) for each toxicity grade supplied in `var`.
#'
#' For each combination of `param` and `grade_dir` levels, patient counts by worst
#' grade are calculated as follows:
#'   * `1` to `4`: The number of patients with worst grades 1-4, respectively.
#'   * `Any`: The number of patients with at least one abnormality (i.e. grade is not 0).
#'
#' Fractions are calculated by dividing the above counts by the number of patients with at least one
#' valid measurement recorded during treatment.
#'
#' Pre-processing is crucial when using this function and can be done automatically using the
#' [h_adlb_abnormal_by_worst_grade()] helper function. See the description of this function for details on the
#' necessary pre-processing steps.
#'
#' Prior to using this function in your table layout you must use [rtables::split_rows_by()] to create two row
#' splits, one on variable `param` and one on variable `grade_dir`.
#'
#' @inheritParams argument_convention
#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("abnormal_by_worst_grade")`
#'   to see available statistics for this function.
#'
#' @seealso [h_adlb_abnormal_by_worst_grade()] which pre-processes ADLB data frames to be used in
#'   [count_abnormal_by_worst_grade()].
#'
#' @name abnormal_by_worst_grade
#' @order 1
NULL

#' @describeIn abnormal_by_worst_grade Statistics function which counts patients by worst grade.
#'
#' @return
#' * `s_count_abnormal_by_worst_grade()` returns the single statistic `count_fraction` with grades 1 to 4 and
#'   "Any" results.
#'
#' @keywords internal
s_count_abnormal_by_worst_grade <- function(df, # nolint
                                            .var = "GRADE_ANL",
                                            .spl_context,
                                            variables = list(
                                              id = "USUBJID",
                                              param = "PARAM",
                                              grade_dir = "GRADE_DIR"
                                            )) {
  checkmate::assert_string(.var)
  assert_valid_factor(df[[.var]])
  assert_valid_factor(df[[variables$param]])
  assert_valid_factor(df[[variables$grade_dir]])
  assert_df_with_variables(df, c(a = .var, variables))
  checkmate::assert_multi_class(df[[variables$id]], classes = c("factor", "character"))

  # To verify that the `split_rows_by` are performed with correct variables.
  checkmate::assert_subset(c(variables[["param"]], variables[["grade_dir"]]), .spl_context$split)
  first_row <- .spl_context[.spl_context$split == variables[["param"]], ]
  x_lvls <- c(setdiff(levels(df[[.var]]), "0"), "Any")
  result <- split(numeric(0), factor(x_lvls))

  subj <- first_row$full_parent_df[[1]][[variables[["id"]]]]
  subj_cur_col <- subj[first_row$cur_col_subset[[1]]]
  # Some subjects may have a record for high and low directions but
  # should be counted only once.
  denom <- length(unique(subj_cur_col))

  for (lvl in x_lvls) {
    if (lvl != "Any") {
      df_lvl <- df[df[[.var]] == lvl, ]
    } else {
      df_lvl <- df[df[[.var]] != 0, ]
    }
    num <- length(unique(df_lvl[[variables[["id"]]]]))
    fraction <- ifelse(denom == 0, 0, num / denom)
    result[[lvl]] <- formatters::with_label(c(count = num, fraction = fraction), lvl)
  }

  result <- list(count_fraction = result)
  result
}

#' @describeIn abnormal_by_worst_grade Formatted analysis function which is used as `afun`
#'   in `count_abnormal_by_worst_grade()`.
#'
#' @return
#' * `a_count_abnormal_by_worst_grade()` returns the corresponding list with formatted [rtables::CellValue()].
#'
#' @keywords internal
a_count_abnormal_by_worst_grade <- make_afun( # nolint
  s_count_abnormal_by_worst_grade,
  .formats = c(count_fraction = format_count_fraction)
)

#' @describeIn abnormal_by_worst_grade Layout-creating function which can take statistics function arguments
#'   and additional format arguments. This function is a wrapper for [rtables::analyze()].
#'
#' @return
#' * `count_abnormal_by_worst_grade()` returns a layout object suitable for passing to further layouting functions,
#'   or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
#'   the statistics from `s_count_abnormal_by_worst_grade()` to the table layout.
#'
#' @examples
#' library(dplyr)
#' library(forcats)
#' adlb <- tern_ex_adlb
#'
#' # Data is modified in order to have some parameters with grades only in one direction
#' # and simulate the real data.
#' adlb$ATOXGR[adlb$PARAMCD == "ALT" & adlb$ATOXGR %in% c("1", "2", "3", "4")] <- "-1"
#' adlb$ANRIND[adlb$PARAMCD == "ALT" & adlb$ANRIND == "HIGH"] <- "LOW"
#' adlb$WGRHIFL[adlb$PARAMCD == "ALT"] <- ""
#'
#' adlb$ATOXGR[adlb$PARAMCD == "IGA" & adlb$ATOXGR %in% c("-1", "-2", "-3", "-4")] <- "1"
#' adlb$ANRIND[adlb$PARAMCD == "IGA" & adlb$ANRIND == "LOW"] <- "HIGH"
#' adlb$WGRLOFL[adlb$PARAMCD == "IGA"] <- ""
#'
#' # Pre-processing
#' adlb_f <- adlb %>% h_adlb_abnormal_by_worst_grade()
#'
#' # Map excludes records without abnormal grade since they should not be displayed
#' # in the table.
#' map <- unique(adlb_f[adlb_f$GRADE_DIR != "ZERO", c("PARAM", "GRADE_DIR", "GRADE_ANL")]) %>%
#'   lapply(as.character) %>%
#'   as.data.frame() %>%
#'   arrange(PARAM, desc(GRADE_DIR), GRADE_ANL)
#'
#' basic_table() %>%
#'   split_cols_by("ARMCD") %>%
#'   split_rows_by("PARAM") %>%
#'   split_rows_by("GRADE_DIR", split_fun = trim_levels_to_map(map)) %>%
#'   count_abnormal_by_worst_grade(
#'     var = "GRADE_ANL",
#'     variables = list(id = "USUBJID", param = "PARAM", grade_dir = "GRADE_DIR")
#'   ) %>%
#'   build_table(df = adlb_f)
#'
#' @export
#' @order 2
count_abnormal_by_worst_grade <- function(lyt,
                                          var,
                                          variables = list(
                                            id = "USUBJID",
                                            param = "PARAM",
                                            grade_dir = "GRADE_DIR"
                                          ),
                                          na_str = default_na_str(),
                                          nested = TRUE,
                                          ...,
                                          .stats = NULL,
                                          .formats = NULL,
                                          .labels = NULL,
                                          .indent_mods = NULL) {
  extra_args <- list(variables = variables, ...)

  afun <- make_afun(
    a_count_abnormal_by_worst_grade,
    .stats = .stats,
    .formats = .formats,
    .labels = .labels,
    .indent_mods = .indent_mods,
    .ungroup_stats = "count_fraction"
  )
  analyze(
    lyt = lyt,
    vars = var,
    afun = afun,
    na_str = na_str,
    nested = nested,
    extra_args = extra_args,
    show_labels = "hidden"
  )
}

#' Helper function to prepare ADLB for `count_abnormal_by_worst_grade()`
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Helper function to prepare an ADLB data frame to be used as input in
#' [count_abnormal_by_worst_grade()]. The following pre-processing steps are applied:
#'
#' 1. `adlb` is filtered on variable `avisit` to only include post-baseline visits.
#' 2. `adlb` is filtered on variables `worst_flag_low` and `worst_flag_high` so that only
#'    worst grades (in either direction) are included.
#' 3. From the standard lab grade variable `atoxgr`, the following two variables are derived
#'    and added to `adlb`:
#'   * A grade direction variable (e.g. `GRADE_DIR`). The variable takes value `"HIGH"` when
#'     `atoxgr > 0`, `"LOW"` when `atoxgr < 0`, and `"ZERO"` otherwise.
#'   * A toxicity grade variable (e.g. `GRADE_ANL`) where all negative values from `atoxgr` are
#'     replaced by their absolute values.
#' 4. Unused factor levels are dropped from `adlb` via [droplevels()].
#'
#' @param adlb (`data.frame`)\cr ADLB data frame.
#' @param atoxgr (`string`)\cr name of the analysis toxicity grade variable. This must be a `factor`
#'   variable.
#' @param avisit (`string`)\cr name of the analysis visit variable.
#' @param worst_flag_low (`string`)\cr name of the worst low lab grade flag variable. This variable is
#'   set to `"Y"` when indicating records of worst low lab grades.
#' @param worst_flag_high (`string`)\cr name of the worst high lab grade flag variable. This variable is
#'   set to `"Y"` when indicating records of worst high lab grades.
#'
#' @return `h_adlb_abnormal_by_worst_grade()` returns the `adlb` data frame with two new
#'   variables: `GRADE_DIR` and `GRADE_ANL`.
#'
#' @seealso [abnormal_by_worst_grade]
#'
#' @examples
#' h_adlb_abnormal_by_worst_grade(tern_ex_adlb) %>%
#'   dplyr::select(ATOXGR, GRADE_DIR, GRADE_ANL) %>%
#'   head(10)
#'
#' @export
h_adlb_abnormal_by_worst_grade <- function(adlb,
                                           atoxgr = "ATOXGR",
                                           avisit = "AVISIT",
                                           worst_flag_low = "WGRLOFL",
                                           worst_flag_high = "WGRHIFL") {
  adlb %>%
    dplyr::filter(
      !.data[[avisit]] %in% c("SCREENING", "BASELINE"),
      .data[[worst_flag_low]] == "Y" | .data[[worst_flag_high]] == "Y"
    ) %>%
    dplyr::mutate(
      GRADE_DIR = factor(
        dplyr::case_when(
          .data[[atoxgr]] %in% c("-1", "-2", "-3", "-4") ~ "LOW",
          .data[[atoxgr]] == "0" ~ "ZERO",
          .data[[atoxgr]] %in% c("1", "2", "3", "4") ~ "HIGH"
        ),
        levels = c("LOW", "ZERO", "HIGH")
      ),
      GRADE_ANL = forcats::fct_relevel(
        forcats::fct_recode(.data[[atoxgr]], `1` = "-1", `2` = "-2", `3` = "-3", `4` = "-4"),
        c("0", "1", "2", "3", "4")
      )
    ) %>%
    droplevels()
}

#' Count patients with abnormal range values
#'
#' @description `r lifecycle::badge("stable")`
#'
#' The analyze function [count_abnormal()] creates a layout element to count patients with abnormal analysis range
#' values in each direction.
#'
#' This function analyzes primary analysis variable `var` which indicates abnormal range results.
#' Additional analysis variables that can be supplied as a list via the `variables` parameter are
#' `id` (defaults to `USUBJID`), a variable to indicate unique subject identifiers, and `baseline`
#' (defaults to `BNRIND`), a variable to indicate baseline reference ranges.
#'
#' For each direction specified via the `abnormal` parameter (e.g. High or Low), a fraction of
#' patient counts is returned, with numerator and denominator calculated as follows:
#'   * `num`: The number of patients with this abnormality recorded while on treatment.
#'   * `denom`: The total number of patients with at least one post-baseline assessment.
#'
#' This function assumes that `df` has been filtered to only include post-baseline records.
#'
#' @inheritParams argument_convention
#' @param abnormal (named `list`)\cr list identifying the abnormal range level(s) in `var`. Defaults to
#'   `list(Low = "LOW", High = "HIGH")` but you can also group different levels into the named list,
#'   for example, `abnormal = list(Low = c("LOW", "LOW LOW"), High = c("HIGH", "HIGH HIGH"))`.
#' @param exclude_base_abn (`flag`)\cr whether to exclude subjects with baseline abnormality
#'   from numerator and denominator.
#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("abnormal")`
#'   to see available statistics for this function.
#'
#' @note
#' * `count_abnormal()` only considers a single variable that contains multiple abnormal levels.
#' * `df` should be filtered to only include post-baseline records.
#' * The denominator includes patients that may have other abnormal levels at baseline,
#'   and patients missing baseline records. Patients with these abnormalities at
#'   baseline can be optionally excluded from numerator and denominator via the
#'   `exclude_base_abn` parameter.
#'
#' @name abnormal
#' @include formatting_functions.R
#' @order 1
NULL

#' @describeIn abnormal Statistics function which counts patients with abnormal range values
#'   for a single `abnormal` level.
#'
#' @return
#' * `s_count_abnormal()` returns the statistic `fraction` which is a vector with `num` and `denom` counts of patients.
#'
#' @keywords internal
s_count_abnormal <- function(df,
                             .var,
                             abnormal = list(Low = "LOW", High = "HIGH"),
                             variables = list(id = "USUBJID", baseline = "BNRIND"),
                             exclude_base_abn = FALSE) {
  checkmate::assert_list(abnormal, types = "character", names = "named", len = 2, any.missing = FALSE)
  checkmate::assert_true(any(unlist(abnormal) %in% levels(df[[.var]])))
  checkmate::assert_factor(df[[.var]])
  checkmate::assert_flag(exclude_base_abn)
  assert_df_with_variables(df, c(range = .var, variables))
  checkmate::assert_multi_class(df[[variables$baseline]], classes = c("factor", "character"))
  checkmate::assert_multi_class(df[[variables$id]], classes = c("factor", "character"))

  count_abnormal_single <- function(abn_name, abn) {
    # Patients in the denominator fulfill:
    # - have at least one post-baseline visit
    # - their baseline must not be abnormal if `exclude_base_abn`.
    if (exclude_base_abn) {
      denom_select <- !(df[[variables$baseline]] %in% abn)
    } else {
      denom_select <- TRUE
    }
    denom <- length(unique(df[denom_select, variables$id, drop = TRUE]))

    # Patients in the numerator fulfill:
    # - have at least one post-baseline visit with the required abnormality level
    # - are part of the denominator patients.
    num_select <- (df[[.var]] %in% abn) & denom_select
    num <- length(unique(df[num_select, variables$id, drop = TRUE]))

    formatters::with_label(c(num = num, denom = denom), abn_name)
  }

  # This will define the abnormal levels theoretically possible for a specific lab parameter
  # within a split level of a layout.
  abnormal_lev <- lapply(abnormal, intersect, levels(df[[.var]]))
  abnormal_lev <- abnormal_lev[vapply(abnormal_lev, function(x) length(x) > 0, logical(1))]

  result <- sapply(names(abnormal_lev), function(i) count_abnormal_single(i, abnormal_lev[[i]]), simplify = FALSE)
  result <- list(fraction = result)
  result
}

#' @describeIn abnormal Formatted analysis function which is used as `afun` in `count_abnormal()`.
#'
#' @return
#' * `a_count_abnormal()` returns the corresponding list with formatted [rtables::CellValue()].
#'
#' @keywords internal
a_count_abnormal <- make_afun(
  s_count_abnormal,
  .formats = c(fraction = format_fraction)
)

#' @describeIn abnormal Layout-creating function which can take statistics function arguments
#'   and additional format arguments. This function is a wrapper for [rtables::analyze()].
#'
#' @return
#' * `count_abnormal()` returns a layout object suitable for passing to further layouting functions,
#'   or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
#'   the statistics from `s_count_abnormal()` to the table layout.
#'
#' @examples
#' library(dplyr)
#'
#' df <- data.frame(
#'   USUBJID = as.character(c(1, 1, 2, 2)),
#'   ANRIND = factor(c("NORMAL", "LOW", "HIGH", "HIGH")),
#'   BNRIND = factor(c("NORMAL", "NORMAL", "HIGH", "HIGH")),
#'   ONTRTFL = c("", "Y", "", "Y"),
#'   stringsAsFactors = FALSE
#' )
#'
#' # Select only post-baseline records.
#' df <- df %>%
#'   filter(ONTRTFL == "Y")
#'
#' # Layout creating function.
#' basic_table() %>%
#'   count_abnormal(var = "ANRIND", abnormal = list(high = "HIGH", low = "LOW")) %>%
#'   build_table(df)
#'
#' # Passing of statistics function and formatting arguments.
#' df2 <- data.frame(
#'   ID = as.character(c(1, 1, 2, 2)),
#'   RANGE = factor(c("NORMAL", "LOW", "HIGH", "HIGH")),
#'   BL_RANGE = factor(c("NORMAL", "NORMAL", "HIGH", "HIGH")),
#'   ONTRTFL = c("", "Y", "", "Y"),
#'   stringsAsFactors = FALSE
#' )
#'
#' # Select only post-baseline records.
#' df2 <- df2 %>%
#'   filter(ONTRTFL == "Y")
#'
#' basic_table() %>%
#'   count_abnormal(
#'     var = "RANGE",
#'     abnormal = list(low = "LOW", high = "HIGH"),
#'     variables = list(id = "ID", baseline = "BL_RANGE")
#'   ) %>%
#'   build_table(df2)
#'
#' @export
#' @order 2
count_abnormal <- function(lyt,
                           var,
                           abnormal = list(Low = "LOW", High = "HIGH"),
                           variables = list(id = "USUBJID", baseline = "BNRIND"),
                           exclude_base_abn = FALSE,
                           na_str = default_na_str(),
                           nested = TRUE,
                           ...,
                           table_names = var,
                           .stats = NULL,
                           .formats = NULL,
                           .labels = NULL,
                           .indent_mods = NULL) {
  extra_args <- list(abnormal = abnormal, variables = variables, exclude_base_abn = exclude_base_abn, ...)

  afun <- make_afun(
    a_count_abnormal,
    .stats = .stats,
    .formats = .formats,
    .labels = .labels,
    .indent_mods = .indent_mods,
    .ungroup_stats = "fraction"
  )

  checkmate::assert_string(var)

  analyze(
    lyt = lyt,
    vars = var,
    afun = afun,
    na_str = na_str,
    nested = nested,
    table_names = table_names,
    extra_args = extra_args,
    show_labels = "hidden"
  )
}

#' Compare variables between groups
#'
#' @description `r lifecycle::badge("stable")`
#'
#' The analyze function [compare_vars()] creates a layout element to summarize and compare one or more variables, using
#' the S3 generic function [s_summary()] to calculate a list of summary statistics. A list of all available statistics
#' for numeric variables can be viewed by running `get_stats("analyze_vars_numeric", add_pval = TRUE)` and for
#' non-numeric variables by running `get_stats("analyze_vars_counts", add_pval = TRUE)`. Use the `.stats` parameter to
#' specify the statistics to include in your output summary table.
#'
#' Prior to using this function in your table layout you must use [rtables::split_cols_by()] to create a column
#' split on the variable to be used in comparisons, and specify a reference group via the `ref_group` parameter.
#' Comparisons can be performed for each group (column) against the specified reference group by including the p-value
#' statistic.
#'
#' @inheritParams argument_convention
#' @param .stats (`character`)\cr statistics to select for the table. Run
#'   `get_stats("analyze_vars_numeric", add_pval = TRUE)` to see statistics available for numeric variables, and
#'   `get_stats("analyze_vars_counts", add_pval = TRUE)` for statistics available for non-numeric variables.
#'
#' @note
#' * For factor variables, `denom` for factor proportions can only be `n` since the purpose is to compare proportions
#'   between columns, therefore a row-based proportion would not make sense. Proportion based on `N_col` would
#'   be difficult since we use counts for the chi-squared test statistic, therefore missing values should be accounted
#'   for as explicit factor levels.
#' * If factor variables contain `NA`, these `NA` values are excluded by default. To include `NA` values
#'   set `na.rm = FALSE` and missing values will be displayed as an `NA` level. Alternatively, an explicit
#'   factor level can be defined for `NA` values during pre-processing via [df_explicit_na()] - the
#'   default `na_level` (`"<Missing>"`) will also be excluded when `na.rm` is set to `TRUE`.
#' * For character variables, automatic conversion to factor does not guarantee that the table
#'   will be generated correctly. In particular for sparse tables this very likely can fail.
#'   Therefore it is always better to manually convert character variables to factors during pre-processing.
#' * For `compare_vars()`, the column split must define a reference group via `ref_group` so that the comparison
#'   is well defined.
#'
#' @seealso [s_summary()] which is used internally to compute a summary within `s_compare()`, and [a_summary()]
#'   which is used (with `compare = TRUE`) as the analysis function for `compare_vars()`.
#'
#' @name compare_variables
#' @include analyze_variables.R
#' @order 1
NULL

#' @describeIn compare_variables S3 generic function to produce a comparison summary.
#'
#' @return
#' * `s_compare()` returns output of [s_summary()] and comparisons versus the reference group in the form of p-values.
#'
#' @export
s_compare <- function(x,
                      .ref_group,
                      .in_ref_col,
                      ...) {
  UseMethod("s_compare", x)
}

#' @describeIn compare_variables Method for `numeric` class. This uses the standard t-test
#'   to calculate the p-value.
#'
#' @method s_compare numeric
#'
#' @examples
#' # `s_compare.numeric`
#'
#' ## Usual case where both this and the reference group vector have more than 1 value.
#' s_compare(rnorm(10, 5, 1), .ref_group = rnorm(5, -5, 1), .in_ref_col = FALSE)
#'
#' ## If one group has not more than 1 value, then p-value is not calculated.
#' s_compare(rnorm(10, 5, 1), .ref_group = 1, .in_ref_col = FALSE)
#'
#' ## Empty numeric does not fail, it returns NA-filled items and no p-value.
#' s_compare(numeric(), .ref_group = numeric(), .in_ref_col = FALSE)
#'
#' @export
s_compare.numeric <- function(x,
                              .ref_group,
                              .in_ref_col,
                              ...) {
  checkmate::assert_numeric(x)
  checkmate::assert_numeric(.ref_group)
  checkmate::assert_flag(.in_ref_col)

  y <- s_summary.numeric(x = x, ...)

  y$pval <- if (!.in_ref_col && n_available(x) > 1 && n_available(.ref_group) > 1) {
    stats::t.test(x, .ref_group)$p.value
  } else {
    character()
  }

  y
}

#' @describeIn compare_variables Method for `factor` class. This uses the chi-squared test
#'   to calculate the p-value.
#'
#' @param denom (`string`)\cr choice of denominator for factor proportions,
#'   can only be `n` (number of values in this row and column intersection).
#'
#' @method s_compare factor
#'
#' @examples
#' # `s_compare.factor`
#'
#' ## Basic usage:
#' x <- factor(c("a", "a", "b", "c", "a"))
#' y <- factor(c("a", "b", "c"))
#' s_compare(x = x, .ref_group = y, .in_ref_col = FALSE)
#'
#' ## Management of NA values.
#' x <- explicit_na(factor(c("a", "a", "b", "c", "a", NA, NA)))
#' y <- explicit_na(factor(c("a", "b", "c", NA)))
#' s_compare(x = x, .ref_group = y, .in_ref_col = FALSE, na.rm = TRUE)
#' s_compare(x = x, .ref_group = y, .in_ref_col = FALSE, na.rm = FALSE)
#'
#' @export
s_compare.factor <- function(x,
                             .ref_group,
                             .in_ref_col,
                             denom = "n",
                             na.rm = TRUE, # nolint
                             ...) {
  checkmate::assert_flag(.in_ref_col)
  assert_valid_factor(x)
  assert_valid_factor(.ref_group)
  denom <- match.arg(denom)

  y <- s_summary.factor(
    x = x,
    denom = denom,
    na.rm = na.rm,
    ...
  )

  if (na.rm) {
    x <- x[!is.na(x)] %>% fct_discard("<Missing>")
    .ref_group <- .ref_group[!is.na(.ref_group)] %>% fct_discard("<Missing>")
  } else {
    x <- x %>% explicit_na(label = "NA")
    .ref_group <- .ref_group %>% explicit_na(label = "NA")
  }

  if ("NA" %in% levels(x)) levels(.ref_group) <- c(levels(.ref_group), "NA")
  checkmate::assert_factor(x, levels = levels(.ref_group), min.levels = 2)

  y$pval_counts <- if (!.in_ref_col && length(x) > 0 && length(.ref_group) > 0) {
    tab <- rbind(table(x), table(.ref_group))
    res <- suppressWarnings(stats::chisq.test(tab))
    res$p.value
  } else {
    character()
  }

  y
}

#' @describeIn compare_variables Method for `character` class. This makes an automatic
#'   conversion to `factor` (with a warning) and then forwards to the method for factors.
#'
#' @param verbose (`flag`)\cr whether warnings and messages should be printed. Mainly used
#'   to print out information about factor casting. Defaults to `TRUE`.
#'
#' @method s_compare character
#'
#' @examples
#' # `s_compare.character`
#'
#' ## Basic usage:
#' x <- c("a", "a", "b", "c", "a")
#' y <- c("a", "b", "c")
#' s_compare(x, .ref_group = y, .in_ref_col = FALSE, .var = "x", verbose = FALSE)
#'
#' ## Note that missing values handling can make a large difference:
#' x <- c("a", "a", "b", "c", "a", NA)
#' y <- c("a", "b", "c", rep(NA, 20))
#' s_compare(x,
#'   .ref_group = y, .in_ref_col = FALSE,
#'   .var = "x", verbose = FALSE
#' )
#' s_compare(x,
#'   .ref_group = y, .in_ref_col = FALSE, .var = "x",
#'   na.rm = FALSE, verbose = FALSE
#' )
#'
#' @export
s_compare.character <- function(x,
                                .ref_group,
                                .in_ref_col,
                                denom = "n",
                                na.rm = TRUE, # nolint
                                .var,
                                verbose = TRUE,
                                ...) {
  x <- as_factor_keep_attributes(x, verbose = verbose)
  .ref_group <- as_factor_keep_attributes(.ref_group, verbose = verbose)
  s_compare(
    x = x,
    .ref_group = .ref_group,
    .in_ref_col = .in_ref_col,
    denom = denom,
    na.rm = na.rm,
    ...
  )
}

#' @describeIn compare_variables Method for `logical` class. A chi-squared test
#'   is used. If missing values are not removed, then they are counted as `FALSE`.
#'
#' @method s_compare logical
#'
#' @examples
#' # `s_compare.logical`
#'
#' ## Basic usage:
#' x <- c(TRUE, FALSE, TRUE, TRUE)
#' y <- c(FALSE, FALSE, TRUE)
#' s_compare(x, .ref_group = y, .in_ref_col = FALSE)
#'
#' ## Management of NA values.
#' x <- c(NA, TRUE, FALSE)
#' y <- c(NA, NA, NA, NA, FALSE)
#' s_compare(x, .ref_group = y, .in_ref_col = FALSE, na.rm = TRUE)
#' s_compare(x, .ref_group = y, .in_ref_col = FALSE, na.rm = FALSE)
#'
#' @export
s_compare.logical <- function(x,
                              .ref_group,
                              .in_ref_col,
                              na.rm = TRUE, # nolint
                              denom = "n",
                              ...) {
  denom <- match.arg(denom)

  y <- s_summary.logical(
    x = x,
    na.rm = na.rm,
    denom = denom,
    ...
  )

  if (na.rm) {
    x <- stats::na.omit(x)
    .ref_group <- stats::na.omit(.ref_group)
  } else {
    x[is.na(x)] <- FALSE
    .ref_group[is.na(.ref_group)] <- FALSE
  }

  y$pval_counts <- if (!.in_ref_col && length(x) > 0 && length(.ref_group) > 0) {
    x <- factor(x, levels = c(TRUE, FALSE))
    .ref_group <- factor(.ref_group, levels = c(TRUE, FALSE))
    tbl <- rbind(table(x), table(.ref_group))
    suppressWarnings(prop_chisq(tbl))
  } else {
    character()
  }

  y
}

#' @describeIn compare_variables Layout-creating function which can take statistics function arguments
#'   and additional format arguments. This function is a wrapper for [rtables::analyze()].
#'
#' @param ... arguments passed to `s_compare()`.
#' @param .indent_mods (named `integer`)\cr indent modifiers for the labels. Each element of the vector
#'   should be a name-value pair with name corresponding to a statistic specified in `.stats` and value the indentation
#'   for that statistic's row label.
#'
#' @return
#' * `compare_vars()` returns a layout object suitable for passing to further layouting functions,
#'   or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
#'   the statistics from `s_compare()` to the table layout.
#'
#' @examples
#' # `compare_vars()` in `rtables` pipelines
#'
#' ## Default output within a `rtables` pipeline.
#' lyt <- basic_table() %>%
#'   split_cols_by("ARMCD", ref_group = "ARM B") %>%
#'   compare_vars(c("AGE", "SEX"))
#' build_table(lyt, tern_ex_adsl)
#'
#' ## Select and format statistics output.
#' lyt <- basic_table() %>%
#'   split_cols_by("ARMCD", ref_group = "ARM C") %>%
#'   compare_vars(
#'     vars = "AGE",
#'     .stats = c("mean_sd", "pval"),
#'     .formats = c(mean_sd = "xx.x, xx.x"),
#'     .labels = c(mean_sd = "Mean, SD")
#'   )
#' build_table(lyt, df = tern_ex_adsl)
#'
#' @export
#' @order 2
compare_vars <- function(lyt,
                         vars,
                         var_labels = vars,
                         na_str = default_na_str(),
                         nested = TRUE,
                         ...,
                         na.rm = TRUE, # nolint
                         show_labels = "default",
                         table_names = vars,
                         section_div = NA_character_,
                         .stats = c("n", "mean_sd", "count_fraction", "pval"),
                         .formats = NULL,
                         .labels = NULL,
                         .indent_mods = NULL) {
  extra_args <- list(.stats = .stats, na.rm = na.rm, na_str = na_str, compare = TRUE, ...)

  if (!is.null(.formats)) extra_args[[".formats"]] <- .formats
  if (!is.null(.labels)) extra_args[[".labels"]] <- .labels
  if (!is.null(.indent_mods)) extra_args[[".indent_mods"]] <- .indent_mods

  analyze(
    lyt = lyt,
    vars = vars,
    var_labels = var_labels,
    afun = a_summary,
    na_str = na_str,
    nested = nested,
    extra_args = extra_args,
    inclNAs = TRUE,
    show_labels = show_labels,
    table_names = table_names,
    section_div = section_div
  )
}

#' Incidence rate estimation
#'
#' @description `r lifecycle::badge("stable")`
#'
#' The analyze function [estimate_incidence_rate()] creates a layout element to estimate an event rate adjusted for
#' person-years at risk, otherwise known as incidence rate. The primary analysis variable specified via `vars` is
#' the person-years at risk. In addition to this variable, the `n_events` variable for number of events observed (where
#' a value of 1 means an event was observed and 0 means that no event was observed) must also be specified.
#'
#' @inheritParams argument_convention
#' @param control (`list`)\cr parameters for estimation details, specified by using
#'   the helper function [control_incidence_rate()]. Possible parameter options are:
#'   * `conf_level` (`proportion`)\cr confidence level for the estimated incidence rate.
#'   * `conf_type` (`string`)\cr `normal` (default), `normal_log`, `exact`, or `byar`
#'     for confidence interval type.
#'   * `input_time_unit` (`string`)\cr `day`, `week`, `month`, or `year` (default)
#'     indicating time unit for data input.
#'   * `num_pt_year` (`numeric`)\cr time unit for desired output (in person-years).
#' @param n_events (`string`)\cr name of integer variable indicating whether an event has been observed (1) or not (0).
#' @param id_var (`string`)\cr name of variable used as patient identifier if `"n_unique"` is included in `.stats`.
#'   Defaults to `"USUBJID"`.
#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("estimate_incidence_rate")`
#'   to see available statistics for this function.
#' @param summarize (`flag`)\cr whether the function should act as an analyze function (`summarize = FALSE`), or a
#'   summarize function (`summarize = TRUE`). Defaults to `FALSE`.
#' @param label_fmt (`string`)\cr how labels should be formatted after a row split occurs if `summarize = TRUE`. The
#'   string should use `"%s"` to represent row split levels, and `"%.labels"` to represent labels supplied to the
#'   `.labels` argument. Defaults to `"%s - %.labels"`.
#'
#' @seealso [control_incidence_rate()] and helper functions [h_incidence_rate].
#'
#' @examples
#' df <- data.frame(
#'   USUBJID = as.character(seq(6)),
#'   CNSR = c(0, 1, 1, 0, 0, 0),
#'   AVAL = c(10.1, 20.4, 15.3, 20.8, 18.7, 23.4),
#'   ARM = factor(c("A", "A", "A", "B", "B", "B")),
#'   STRATA1 = factor(c("X", "Y", "Y", "X", "X", "Y"))
#' )
#' df$n_events <- 1 - df$CNSR
#'
#' @name incidence_rate
#' @order 1
NULL

#' @describeIn incidence_rate Statistics function which estimates the incidence rate and the
#'   associated confidence interval.
#'
#' @return
#' * `s_incidence_rate()` returns the following statistics:
#'   - `person_years`: Total person-years at risk.
#'   - `n_events`: Total number of events observed.
#'   - `rate`: Estimated incidence rate.
#'   - `rate_ci`: Confidence interval for the incidence rate.
#'   - `n_unique`: Total number of patients with at least one event observed.
#'   - `n_rate`: Total number of events observed & estimated incidence rate.
#'
#' @keywords internal
s_incidence_rate <- function(df,
                             .var,
                             n_events,
                             is_event = lifecycle::deprecated(),
                             id_var = "USUBJID",
                             control = control_incidence_rate()) {
  if (lifecycle::is_present(is_event)) {
    checkmate::assert_string(is_event)
    lifecycle::deprecate_warn(
      "0.9.6", "s_incidence_rate(is_event)", "s_incidence_rate(n_events)"
    )
    n_events <- is_event
    df[[n_events]] <- as.numeric(df[[is_event]])
  }

  assert_df_with_variables(df, list(tte = .var, n_events = n_events))
  checkmate::assert_string(.var)
  checkmate::assert_string(n_events)
  checkmate::assert_string(id_var)
  checkmate::assert_numeric(df[[.var]], any.missing = FALSE)
  checkmate::assert_integerish(df[[n_events]], any.missing = FALSE)

  n_unique <- n_available(unique(df[[id_var]][df[[n_events]] == 1]))
  input_time_unit <- control$input_time_unit
  num_pt_year <- control$num_pt_year
  conf_level <- control$conf_level
  person_years <- sum(df[[.var]], na.rm = TRUE) * (
    1 * (input_time_unit == "year") +
      1 / 12 * (input_time_unit == "month") +
      1 / 52.14 * (input_time_unit == "week") +
      1 / 365.24 * (input_time_unit == "day")
  )
  n_events <- sum(df[[n_events]], na.rm = TRUE)

  result <- h_incidence_rate(
    person_years,
    n_events,
    control
  )
  list(
    person_years = formatters::with_label(person_years, "Total patient-years at risk"),
    n_events = formatters::with_label(n_events, "Number of adverse events observed"),
    rate = formatters::with_label(result$rate, paste("AE rate per", num_pt_year, "patient-years")),
    rate_ci = formatters::with_label(result$rate_ci, f_conf_level(conf_level)),
    n_unique = formatters::with_label(n_unique, "Total number of patients with at least one adverse event"),
    n_rate = formatters::with_label(
      c(n_events, result$rate),
      paste("Number of adverse events observed (AE rate per", num_pt_year, "patient-years)")
    )
  )
}

#' @describeIn incidence_rate Formatted analysis function which is used as `afun` in `estimate_incidence_rate()`.
#'
#' @return
#' * `a_incidence_rate()` returns the corresponding list with formatted [rtables::CellValue()].
#'
#' @examples
#' a_incidence_rate(
#'   df,
#'   .var = "AVAL",
#'   .df_row = df,
#'   n_events = "n_events"
#' )
#'
#' @export
a_incidence_rate <- function(df,
                             labelstr = "",
                             .var,
                             .df_row,
                             n_events,
                             id_var = "USUBJID",
                             control = control_incidence_rate(),
                             .stats = NULL,
                             .formats = c(
                               "person_years" = "xx.x",
                               "n_events" = "xx",
                               "rate" = "xx.xx",
                               "rate_ci" = "(xx.xx, xx.xx)",
                               "n_unique" = "xx",
                               "n_rate" = "xx (xx.x)"
                             ),
                             .labels = NULL,
                             .indent_mods = NULL,
                             na_str = default_na_str(),
                             label_fmt = "%s - %.labels") {
  checkmate::assert_string(label_fmt)

  x_stats <- s_incidence_rate(
    df = df, .var = .var, n_events = n_events, id_var = id_var, control = control
  )
  if (is.null(unlist(x_stats))) {
    return(NULL)
  }

  # Fill in with defaults
  formats_def <- formals()$.formats %>% eval()
  .formats <- c(.formats, formats_def)[!duplicated(names(c(.formats, formats_def)))]
  labels_def <- sapply(x_stats, \(x) attributes(x)$label)
  .labels <- c(.labels, labels_def)[!duplicated(names(c(.labels, labels_def)))]
  if (nzchar(labelstr) > 0) {
    .labels <- sapply(.labels, \(x) gsub("%.labels", x, gsub("%s", labelstr, label_fmt)))
  }

  # Fill in with formatting defaults if needed
  .stats <- get_stats("estimate_incidence_rate", stats_in = .stats)
  .formats <- get_formats_from_stats(.stats, .formats)
  .labels <- get_labels_from_stats(.stats, .labels)
  .indent_mods <- get_indents_from_stats(.stats, .indent_mods)

  x_stats <- x_stats[.stats]

  in_rows(
    .list = x_stats,
    .formats = .formats,
    .labels = .labels,
    .indent_mods = .indent_mods,
    .format_na_strs = na_str
  )
}

#' @describeIn incidence_rate Layout-creating function which can take statistics function arguments
#'   and additional format arguments. This function is a wrapper for [rtables::analyze()].
#'
#' @return
#' * `estimate_incidence_rate()` returns a layout object suitable for passing to further layouting functions,
#'   or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
#'   the statistics from `s_incidence_rate()` to the table layout.
#'
#' @examples
#' basic_table(show_colcounts = TRUE) %>%
#'   split_cols_by("ARM") %>%
#'   estimate_incidence_rate(
#'     vars = "AVAL",
#'     n_events = "n_events",
#'     control = control_incidence_rate(
#'       input_time_unit = "month",
#'       num_pt_year = 100
#'     )
#'   ) %>%
#'   build_table(df)
#'
#' # summarize = TRUE
#' basic_table(show_colcounts = TRUE) %>%
#'   split_cols_by("ARM") %>%
#'   split_rows_by("STRATA1", child_labels = "visible") %>%
#'   estimate_incidence_rate(
#'     vars = "AVAL",
#'     n_events = "n_events",
#'     .stats = c("n_unique", "n_rate"),
#'     summarize = TRUE,
#'     label_fmt = "%.labels"
#'   ) %>%
#'   build_table(df)
#'
#' @export
#' @order 2
estimate_incidence_rate <- function(lyt,
                                    vars,
                                    n_events,
                                    id_var = "USUBJID",
                                    control = control_incidence_rate(),
                                    na_str = default_na_str(),
                                    nested = TRUE,
                                    summarize = FALSE,
                                    label_fmt = "%s - %.labels",
                                    ...,
                                    show_labels = "hidden",
                                    table_names = vars,
                                    .stats = c("person_years", "n_events", "rate", "rate_ci"),
                                    .formats = NULL,
                                    .labels = NULL,
                                    .indent_mods = NULL) {
  extra_args <- c(
    list(.stats = .stats, .formats = .formats, .labels = .labels, .indent_mods = .indent_mods, na_str = na_str),
    list(n_events = n_events, id_var = id_var, control = control, label_fmt = label_fmt, ...)
  )

  if (!summarize) {
    analyze(
      lyt,
      vars,
      show_labels = show_labels,
      table_names = table_names,
      afun = a_incidence_rate,
      na_str = na_str,
      nested = nested,
      extra_args = extra_args
    )
  } else {
    summarize_row_groups(
      lyt,
      vars,
      cfun = a_incidence_rate,
      na_str = na_str,
      extra_args = extra_args
    )
  }
}

#' Custom split functions
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Collection of useful functions that are expanding on the core list of functions
#' provided by `rtables`. See [rtables::custom_split_funs] and [rtables::make_split_fun()]
#' for more information on how to make a custom split function. All these functions
#' work with [rtables::split_rows_by()] argument `split_fun` to modify the way the split
#' happens. For other split functions, consider consulting [`rtables::split_funcs`].
#'
#' @seealso [rtables::make_split_fun()]
#'
#' @name utils_split_funs
NULL

#' @describeIn utils_split_funs Split function to place reference group facet at a specific position
#'  during post-processing stage.
#'
#' @param position (`string` or `integer`)\cr position to use for the reference group facet. Can be `"first"`,
#'   `"last"`, or a specific position.
#'
#' @return
#' * `ref_group_position()` returns an utility function that puts the reference group
#'   as first, last or at a certain position and needs to be assigned to `split_fun`.
#'
#' @examples
#' library(dplyr)
#'
#' dat <- data.frame(
#'   x = factor(letters[1:5], levels = letters[5:1]),
#'   y = 1:5
#' )
#'
#' # With rtables layout functions
#' basic_table() %>%
#'   split_cols_by("x", ref_group = "c", split_fun = ref_group_position("last")) %>%
#'   analyze("y") %>%
#'   build_table(dat)
#'
#' # With tern layout funcitons
#' adtte_f <- tern_ex_adtte %>%
#'   filter(PARAMCD == "OS") %>%
#'   mutate(
#'     AVAL = day2month(AVAL),
#'     is_event = CNSR == 0
#'   )
#'
#' basic_table() %>%
#'   split_cols_by(var = "ARMCD", ref_group = "ARM B", split_fun = ref_group_position("first")) %>%
#'   add_colcounts() %>%
#'   surv_time(
#'     vars = "AVAL",
#'     var_labels = "Survival Time (Months)",
#'     is_event = "is_event",
#'   ) %>%
#'   build_table(df = adtte_f)
#'
#' basic_table() %>%
#'   split_cols_by(var = "ARMCD", ref_group = "ARM B", split_fun = ref_group_position(2)) %>%
#'   add_colcounts() %>%
#'   surv_time(
#'     vars = "AVAL",
#'     var_labels = "Survival Time (Months)",
#'     is_event = "is_event",
#'   ) %>%
#'   build_table(df = adtte_f)
#'
#' @export
ref_group_position <- function(position = "first") {
  make_split_fun(
    post = list(
      function(splret, spl, fulldf) {
        if (!"ref_group_value" %in% methods::slotNames(spl)) {
          stop("Reference group is undefined.")
        }

        spl_var <- rtables:::spl_payload(spl)
        fulldf[[spl_var]] <- factor(fulldf[[spl_var]])
        init_lvls <- levels(fulldf[[spl_var]])

        if (!all(names(splret$values) %in% init_lvls)) {
          stop("This split function does not work with combination facets.")
        }

        ref_group_pos <- which(init_lvls == rtables:::spl_ref_group(spl))
        pos_choices <- c("first", "last")
        if (checkmate::test_choice(position, pos_choices) && position == "first") {
          pos <- 0
        } else if (checkmate::test_choice(position, pos_choices) && position == "last") {
          pos <- length(init_lvls)
        } else if (checkmate::test_int(position, lower = 1, upper = length(init_lvls))) {
          pos <- position - 1
        } else {
          stop("Wrong input for ref group position. It must be 'first', 'last', or a integer.")
        }

        reord_lvls <- append(init_lvls[-ref_group_pos], init_lvls[ref_group_pos], after = pos)
        ord <- match(reord_lvls, names(splret$values))

        make_split_result(
          splret$values[ord],
          splret$datasplit[ord],
          splret$labels[ord]
        )
      }
    )
  )
}

#' @describeIn utils_split_funs Split function to change level order based on an `integer`
#'   vector or a `character` vector that represent the split variable's factor levels.
#'
#' @param order (`character` or `numeric`)\cr vector of ordering indices for the split facets.
#'
#' @return
#' * `level_order()` returns an utility function that changes the original levels' order,
#'   depending on input `order` and split levels.
#'
#' @examples
#' # level_order --------
#' # Even if default would bring ref_group first, the original order puts it last
#' basic_table() %>%
#'   split_cols_by("Species", split_fun = level_order(c(1, 3, 2))) %>%
#'   analyze("Sepal.Length") %>%
#'   build_table(iris)
#'
#' # character vector
#' new_order <- level_order(levels(iris$Species)[c(1, 3, 2)])
#' basic_table() %>%
#'   split_cols_by("Species", ref_group = "virginica", split_fun = new_order) %>%
#'   analyze("Sepal.Length") %>%
#'   build_table(iris)
#'
#' @export
level_order <- function(order) {
  make_split_fun(
    post = list(
      function(splret, spl, fulldf) {
        if (checkmate::test_integerish(order)) {
          checkmate::assert_integerish(order, lower = 1, upper = length(splret$values))
          ord <- order
        } else {
          checkmate::assert_character(order, len = length(splret$values))
          checkmate::assert_set_equal(order, names(splret$values), ordered = FALSE)
          ord <- match(order, names(splret$values))
        }
        make_split_result(
          splret$values[ord],
          splret$datasplit[ord],
          splret$labels[ord]
        )
      }
    )
  )
}

#' Stack multiple grobs
#'
#' @description `r lifecycle::badge("deprecated")`
#'
#' Stack grobs as a new grob with 1 column and multiple rows layout.
#'
#' @param ... grobs.
#' @param grobs (`list` of `grob`)\cr a list of grobs.
#' @param padding (`grid::unit`)\cr unit of length 1, space between each grob.
#' @param vp (`viewport` or `NULL`)\cr a [viewport()] object (or `NULL`).
#' @param name (`string`)\cr a character identifier for the grob.
#' @param gp (`gpar`)\cr a [gpar()] object.
#'
#' @return A `grob`.
#'
#' @examples
#' library(grid)
#'
#' g1 <- circleGrob(gp = gpar(col = "blue"))
#' g2 <- circleGrob(gp = gpar(col = "red"))
#' g3 <- textGrob("TEST TEXT")
#' grid.newpage()
#' grid.draw(stack_grobs(g1, g2, g3))
#'
#' showViewport()
#'
#' grid.newpage()
#' pushViewport(viewport(layout = grid.layout(1, 2)))
#' vp1 <- viewport(layout.pos.row = 1, layout.pos.col = 2)
#' grid.draw(stack_grobs(g1, g2, g3, vp = vp1, name = "test"))
#'
#' showViewport()
#' grid.ls(grobs = TRUE, viewports = TRUE, print = FALSE)
#'
#' @export
stack_grobs <- function(...,
                        grobs = list(...),
                        padding = grid::unit(2, "line"),
                        vp = NULL,
                        gp = NULL,
                        name = NULL) {
  lifecycle::deprecate_warn(
    "0.9.4",
    "stack_grobs()",
    details = "`tern` plotting functions no longer generate `grob` objects."
  )

  checkmate::assert_true(
    all(vapply(grobs, grid::is.grob, logical(1)))
  )

  if (length(grobs) == 1) {
    return(grobs[[1]])
  }

  n_layout <- 2 * length(grobs) - 1
  hts <- lapply(
    seq(1, n_layout),
    function(i) {
      if (i %% 2 != 0) {
        grid::unit(1, "null")
      } else {
        padding
      }
    }
  )
  hts <- do.call(grid::unit.c, hts)

  main_vp <- grid::viewport(
    layout = grid::grid.layout(nrow = n_layout, ncol = 1, heights = hts)
  )

  nested_grobs <- Map(function(g, i) {
    grid::gTree(
      children = grid::gList(g),
      vp = grid::viewport(layout.pos.row = i, layout.pos.col = 1)
    )
  }, grobs, seq_along(grobs) * 2 - 1)

  grobs_mainvp <- grid::gTree(
    children = do.call(grid::gList, nested_grobs),
    vp = main_vp
  )

  grid::gTree(
    children = grid::gList(grobs_mainvp),
    vp = vp,
    gp = gp,
    name = name
  )
}

#' Arrange multiple grobs
#'
#' @description `r lifecycle::badge("deprecated")`
#'
#' Arrange grobs as a new grob with `n * m (rows * cols)` layout.
#'
#' @inheritParams stack_grobs
#' @param ncol (`integer(1)`)\cr number of columns in layout.
#' @param nrow (`integer(1)`)\cr number of rows in layout.
#' @param padding_ht (`grid::unit`)\cr unit of length 1, vertical space between each grob.
#' @param padding_wt (`grid::unit`)\cr unit of length 1, horizontal space between each grob.
#'
#' @return A `grob`.
#'
#' @examples
#' library(grid)
#'
#' \donttest{
#' num <- lapply(1:9, textGrob)
#' grid::grid.newpage()
#' grid.draw(arrange_grobs(grobs = num, ncol = 2))
#'
#' showViewport()
#'
#' g1 <- circleGrob(gp = gpar(col = "blue"))
#' g2 <- circleGrob(gp = gpar(col = "red"))
#' g3 <- textGrob("TEST TEXT")
#' grid::grid.newpage()
#' grid.draw(arrange_grobs(g1, g2, g3, nrow = 2))
#'
#' showViewport()
#'
#' grid::grid.newpage()
#' grid.draw(arrange_grobs(g1, g2, g3, ncol = 3))
#'
#' grid::grid.newpage()
#' grid::pushViewport(grid::viewport(layout = grid::grid.layout(1, 2)))
#' vp1 <- grid::viewport(layout.pos.row = 1, layout.pos.col = 2)
#' grid.draw(arrange_grobs(g1, g2, g3, ncol = 2, vp = vp1))
#'
#' showViewport()
#' }
#' @export
arrange_grobs <- function(...,
                          grobs = list(...),
                          ncol = NULL, nrow = NULL,
                          padding_ht = grid::unit(2, "line"),
                          padding_wt = grid::unit(2, "line"),
                          vp = NULL,
                          gp = NULL,
                          name = NULL) {
  lifecycle::deprecate_warn(
    "0.9.4",
    "arrange_grobs()",
    details = "`tern` plotting functions no longer generate `grob` objects."
  )

  checkmate::assert_true(
    all(vapply(grobs, grid::is.grob, logical(1)))
  )

  if (length(grobs) == 1) {
    return(grobs[[1]])
  }

  if (is.null(ncol) && is.null(nrow)) {
    ncol <- 1
    nrow <- ceiling(length(grobs) / ncol)
  } else if (!is.null(ncol) && is.null(nrow)) {
    nrow <- ceiling(length(grobs) / ncol)
  } else if (is.null(ncol) && !is.null(nrow)) {
    ncol <- ceiling(length(grobs) / nrow)
  }

  if (ncol * nrow < length(grobs)) {
    stop("specififed ncol and nrow are not enough for arranging the grobs ")
  }

  if (ncol == 1) {
    return(stack_grobs(grobs = grobs, padding = padding_ht, vp = vp, gp = gp, name = name))
  }

  n_col <- 2 * ncol - 1
  n_row <- 2 * nrow - 1
  hts <- lapply(
    seq(1, n_row),
    function(i) {
      if (i %% 2 != 0) {
        grid::unit(1, "null")
      } else {
        padding_ht
      }
    }
  )
  hts <- do.call(grid::unit.c, hts)

  wts <- lapply(
    seq(1, n_col),
    function(i) {
      if (i %% 2 != 0) {
        grid::unit(1, "null")
      } else {
        padding_wt
      }
    }
  )
  wts <- do.call(grid::unit.c, wts)

  main_vp <- grid::viewport(
    layout = grid::grid.layout(nrow = n_row, ncol = n_col, widths = wts, heights = hts)
  )

  nested_grobs <- list()
  k <- 0
  for (i in seq(nrow) * 2 - 1) {
    for (j in seq(ncol) * 2 - 1) {
      k <- k + 1
      if (k <= length(grobs)) {
        nested_grobs <- c(
          nested_grobs,
          list(grid::gTree(
            children = grid::gList(grobs[[k]]),
            vp = grid::viewport(layout.pos.row = i, layout.pos.col = j)
          ))
        )
      }
    }
  }
  grobs_mainvp <- grid::gTree(
    children = do.call(grid::gList, nested_grobs),
    vp = main_vp
  )

  grid::gTree(
    children = grid::gList(grobs_mainvp),
    vp = vp,
    gp = gp,
    name = name
  )
}

#' Draw `grob`
#'
#' @description `r lifecycle::badge("deprecated")`
#'
#' Draw grob on device page.
#'
#' @param grob (`grob`)\cr grid object.
#' @param newpage (`flag`)\cr draw on a new page.
#' @param vp (`viewport` or `NULL`)\cr a [viewport()] object (or `NULL`).
#'
#' @return A `grob`.
#'
#' @examples
#' library(dplyr)
#' library(grid)
#'
#' \donttest{
#' rect <- rectGrob(width = grid::unit(0.5, "npc"), height = grid::unit(0.5, "npc"))
#' rect %>% draw_grob(vp = grid::viewport(angle = 45))
#'
#' num <- lapply(1:10, textGrob)
#' num %>%
#'   arrange_grobs(grobs = .) %>%
#'   draw_grob()
#' showViewport()
#' }
#'
#' @export
draw_grob <- function(grob, newpage = TRUE, vp = NULL) {
  lifecycle::deprecate_warn(
    "0.9.4",
    "draw_grob()",
    details = "`tern` plotting functions no longer generate `grob` objects."
  )

  if (newpage) {
    grid::grid.newpage()
  }
  if (!is.null(vp)) {
    grid::pushViewport(vp)
  }
  grid::grid.draw(grob)
}

tern_grob <- function(x) {
  class(x) <- unique(c("ternGrob", class(x)))
  x
}

#' @keywords internal
print.ternGrob <- function(x, ...) {
  grid::grid.newpage()
  grid::grid.draw(x)
}

#' Estimate proportions of each level of a variable
#'
#' @description `r lifecycle::badge("stable")`
#'
#' The analyze & summarize function [estimate_multinomial_response()] creates a layout element to estimate the
#' proportion and proportion confidence interval for each level of a factor variable. The primary analysis variable,
#' `var`, should be a factor variable, the values of which will be used as labels within the output table.
#'
#' @inheritParams argument_convention
#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("estimate_multinomial_response")`
#'   to see available statistics for this function.
#'
#' @seealso Relevant description function [d_onco_rsp_label()].
#'
#' @name estimate_multinomial_rsp
#' @order 1
NULL

#' Description of standard oncology response
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Describe the oncology response in a standard way.
#'
#' @param x (`character`)\cr the standard oncology codes to be described.
#'
#' @return Response labels.
#'
#' @seealso [estimate_multinomial_rsp()]
#'
#' @examples
#' d_onco_rsp_label(
#'   c("CR", "PR", "SD", "NON CR/PD", "PD", "NE", "Missing", "<Missing>", "NE/Missing")
#' )
#'
#' # Adding some values not considered in d_onco_rsp_label
#'
#' d_onco_rsp_label(
#'   c("CR", "PR", "hello", "hi")
#' )
#'
#' @export
d_onco_rsp_label <- function(x) {
  x <- as.character(x)
  desc <- c(
    CR           = "Complete Response (CR)",
    PR           = "Partial Response (PR)",
    MR           = "Minimal/Minor Response (MR)",
    MRD          = "Minimal Residual Disease (MRD)",
    SD           = "Stable Disease (SD)",
    PD           = "Progressive Disease (PD)",
    `NON CR/PD`  = "Non-CR or Non-PD (NON CR/PD)",
    NE           = "Not Evaluable (NE)",
    `NE/Missing` = "Missing or unevaluable",
    Missing      = "Missing",
    `NA`         = "Not Applicable (NA)",
    ND           = "Not Done (ND)"
  )

  values_label <- vapply(
    X = x,
    FUN.VALUE = character(1),
    function(val) {
      if (val %in% names(desc)) desc[val] else val
    }
  )

  return(factor(values_label, levels = c(intersect(desc, values_label), setdiff(values_label, desc))))
}

#' @describeIn estimate_multinomial_rsp Statistics function which feeds the length of `x` as number
#'   of successes, and `.N_col` as total number of successes and failures into [s_proportion()].
#'
#' @return
#' * `s_length_proportion()` returns statistics from [s_proportion()].
#'
#' @examples
#' s_length_proportion(rep("CR", 10), .N_col = 100)
#' s_length_proportion(factor(character(0)), .N_col = 100)
#'
#' @export
s_length_proportion <- function(x,
                                .N_col, # nolint
                                ...) {
  checkmate::assert_multi_class(x, classes = c("factor", "character"))
  checkmate::assert_vector(x, min.len = 0, max.len = .N_col)
  checkmate::assert_vector(unique(x), min.len = 0, max.len = 1)

  n_true <- length(x)
  n_false <- .N_col - n_true
  x_logical <- rep(c(TRUE, FALSE), c(n_true, n_false))
  s_proportion(df = x_logical, ...)
}

#' @describeIn estimate_multinomial_rsp Formatted analysis function which is used as `afun`
#'   in `estimate_multinomial_response()`.
#'
#' @return
#' * `a_length_proportion()` returns the corresponding list with formatted [rtables::CellValue()].
#'
#' @examples
#' a_length_proportion(rep("CR", 10), .N_col = 100)
#' a_length_proportion(factor(character(0)), .N_col = 100)
#'
#' @export
a_length_proportion <- make_afun(
  s_length_proportion,
  .formats = c(
    n_prop = "xx (xx.x%)",
    prop_ci = "(xx.xx, xx.xx)"
  )
)

#' @describeIn estimate_multinomial_rsp Layout-creating function which can take statistics function arguments
#'   and additional format arguments. This function is a wrapper for [rtables::analyze()] and
#'   [rtables::summarize_row_groups()].
#'
#' @return
#' * `estimate_multinomial_response()` returns a layout object suitable for passing to further layouting functions,
#'   or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
#'   the statistics from `s_length_proportion()` to the table layout.
#'
#' @examples
#' library(dplyr)
#'
#' # Use of the layout creating function.
#' dta_test <- data.frame(
#'   USUBJID = paste0("S", 1:12),
#'   ARM     = factor(rep(LETTERS[1:3], each = 4)),
#'   AVAL    = c(A = c(1, 1, 1, 1), B = c(0, 0, 1, 1), C = c(0, 0, 0, 0))
#' ) %>% mutate(
#'   AVALC = factor(AVAL,
#'     levels = c(0, 1),
#'     labels = c("Complete Response (CR)", "Partial Response (PR)")
#'   )
#' )
#'
#' lyt <- basic_table() %>%
#'   split_cols_by("ARM") %>%
#'   estimate_multinomial_response(var = "AVALC")
#'
#' tbl <- build_table(lyt, dta_test)
#'
#' tbl
#'
#' @export
#' @order 2
estimate_multinomial_response <- function(lyt,
                                          var,
                                          na_str = default_na_str(),
                                          nested = TRUE,
                                          ...,
                                          show_labels = "hidden",
                                          table_names = var,
                                          .stats = "prop_ci",
                                          .formats = NULL,
                                          .labels = NULL,
                                          .indent_mods = NULL) {
  extra_args <- list(...)

  afun <- make_afun(
    a_length_proportion,
    .stats = .stats,
    .formats = .formats,
    .labels = .labels,
    .indent_mods = .indent_mods
  )
  lyt <- split_rows_by(lyt, var = var)
  lyt <- summarize_row_groups(lyt, na_str = na_str)

  analyze(
    lyt,
    vars = var,
    afun = afun,
    show_labels = show_labels,
    table_names = table_names,
    na_str = na_str,
    nested = nested,
    extra_args = extra_args
  )
}

#' Count patient events in columns
#'
#' @description `r lifecycle::badge("stable")`
#'
#' The summarize function [summarize_patients_events_in_cols()] creates a layout element to summarize patient
#' event counts in columns.
#'
#' This function analyzes the elements (events) supplied via the `filters_list` parameter and returns a row
#' with counts of number of patients for each event as well as the total numbers of patients and events.
#' The `id` variable is used to indicate unique subject identifiers (defaults to `USUBJID`).
#'
#' If there are multiple occurrences of the same event recorded for a patient, the event is only counted once.
#'
#' @inheritParams argument_convention
#' @param filters_list (named `list` of `character`)\cr list where each element in this list describes one
#'   type of event describe by filters, in the same format as [s_count_patients_with_event()].
#'   If it has a label, then this will be used for the column title.
#' @param empty_stats (`character`)\cr optional names of the statistics that should be returned empty such
#'   that corresponding table cells will stay blank.
#' @param custom_label (`string` or `NULL`)\cr if provided and `labelstr` is empty then this will
#'   be used as label.
#' @param .stats (`character`)\cr statistics to select for the table. Run
#'   `get_stats("summarize_patients_events_in_cols")` to see available statistics for this function, in addition
#'   to any added using `filters_list`.
#'
#' @name count_patients_events_in_cols
#' @order 1
NULL

#' @describeIn count_patients_events_in_cols Statistics function which counts numbers of patients and multiple
#'   events defined by filters. Used as analysis function `afun` in `summarize_patients_events_in_cols()`.
#'
#' @return
#' * `s_count_patients_and_multiple_events()` returns a list with the statistics:
#'   - `unique`: number of unique patients in `df`.
#'   - `all`: number of rows in `df`.
#'   - one element with the same name as in `filters_list`: number of rows in `df`,
#'     i.e. events, fulfilling the filter condition.
#'
#' @keywords internal
s_count_patients_and_multiple_events <- function(df, # nolint
                                                 id,
                                                 filters_list,
                                                 empty_stats = character(),
                                                 labelstr = "",
                                                 custom_label = NULL) {
  checkmate::assert_list(filters_list, names = "named")
  checkmate::assert_data_frame(df)
  checkmate::assert_string(id)
  checkmate::assert_disjunct(c("unique", "all"), names(filters_list))
  checkmate::assert_character(empty_stats)
  checkmate::assert_string(labelstr)
  checkmate::assert_string(custom_label, null.ok = TRUE)

  # Below we want to count each row in `df` once, therefore introducing this helper index column.
  df$.row_index <- as.character(seq_len(nrow(df)))
  y <- list()
  row_label <- if (labelstr != "") {
    labelstr
  } else if (!is.null(custom_label)) {
    custom_label
  } else {
    "counts"
  }
  y$unique <- formatters::with_label(
    s_num_patients_content(df = df, .N_col = 1, .var = id, required = NULL)$unique[1L],
    row_label
  )
  y$all <- formatters::with_label(
    nrow(df),
    row_label
  )
  events <- Map(
    function(filters) {
      formatters::with_label(
        s_count_patients_with_event(df = df, .var = ".row_index", filters = filters, .N_col = 1, .N_row = 1)$count,
        row_label
      )
    },
    filters = filters_list
  )
  y_complete <- c(y, events)
  y <- if (length(empty_stats) > 0) {
    y_reduced <- y_complete
    for (stat in intersect(names(y_complete), empty_stats)) {
      y_reduced[[stat]] <- formatters::with_label(character(), obj_label(y_reduced[[stat]]))
    }
    y_reduced
  } else {
    y_complete
  }
  y
}

#' @describeIn count_patients_events_in_cols Layout-creating function which can take statistics function
#'   arguments and additional format arguments. This function is a wrapper for [rtables::summarize_row_groups()].
#'
#' @param col_split (`flag`)\cr whether the columns should be split.
#'   Set to `FALSE` when the required column split has been done already earlier in the layout pipe.
#'
#' @return
#' * `summarize_patients_events_in_cols()` returns a layout object suitable for passing to further layouting functions,
#'   or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted content rows
#'   containing the statistics from `s_count_patients_and_multiple_events()` to the table layout.
#'
#' @examples
#' df <- data.frame(
#'   USUBJID = rep(c("id1", "id2", "id3", "id4"), c(2, 3, 1, 1)),
#'   ARM = c("A", "A", "B", "B", "B", "B", "A"),
#'   AESER = rep("Y", 7),
#'   AESDTH = c("Y", "Y", "N", "Y", "Y", "N", "N"),
#'   AEREL = c("Y", "Y", "N", "Y", "Y", "N", "Y"),
#'   AEDECOD = c("A", "A", "A", "B", "B", "C", "D"),
#'   AEBODSYS = rep(c("SOC1", "SOC2", "SOC3"), c(3, 3, 1))
#' )
#'
#' # `summarize_patients_events_in_cols()`
#' basic_table() %>%
#'   summarize_patients_events_in_cols(
#'     filters_list = list(
#'       related = formatters::with_label(c(AEREL = "Y"), "Events (Related)"),
#'       fatal = c(AESDTH = "Y"),
#'       fatal_related = c(AEREL = "Y", AESDTH = "Y")
#'     ),
#'     custom_label = "%s Total number of patients and events"
#'   ) %>%
#'   build_table(df)
#'
#' @export
#' @order 2
summarize_patients_events_in_cols <- function(lyt, # nolint
                                              id = "USUBJID",
                                              filters_list = list(),
                                              empty_stats = character(),
                                              na_str = default_na_str(),
                                              ...,
                                              .stats = c(
                                                "unique",
                                                "all",
                                                names(filters_list)
                                              ),
                                              .labels = c(
                                                unique = "Patients (All)",
                                                all = "Events (All)",
                                                labels_or_names(filters_list)
                                              ),
                                              col_split = TRUE) {
  extra_args <- list(id = id, filters_list = filters_list, empty_stats = empty_stats, ...)

  afun_list <- Map(
    function(stat) {
      make_afun(
        s_count_patients_and_multiple_events,
        .stats = stat,
        .formats = "xx."
      )
    },
    stat = .stats
  )
  if (col_split) {
    lyt <- split_cols_by_multivar(
      lyt = lyt,
      vars = rep(id, length(.stats)),
      varlabels = .labels[.stats]
    )
  }
  summarize_row_groups(
    lyt = lyt,
    cfun = afun_list,
    na_str = na_str,
    extra_args = extra_args
  )
}

#' Sort pharmacokinetic data by `PARAM` variable
#'
#' @description `r lifecycle::badge("stable")`
#'
#' @param pk_data (`data.frame`)\cr pharmacokinetic data frame.
#' @param key_var (`string`)\cr key variable used to merge pk_data and metadata created by [d_pkparam()].
#'
#' @return A pharmacokinetic `data.frame` sorted by a `PARAM` variable.
#'
#' @examples
#' library(dplyr)
#'
#' adpp <- tern_ex_adpp %>% mutate(PKPARAM = factor(paste0(PARAM, " (", AVALU, ")")))
#' pk_ordered_data <- h_pkparam_sort(adpp)
#'
#' @export
h_pkparam_sort <- function(pk_data, key_var = "PARAMCD") {
  assert_df_with_variables(pk_data, list(key_var = key_var))
  pk_data$PARAMCD <- pk_data[[key_var]]

  ordered_pk_data <- d_pkparam()

  # Add the numeric values from ordered_pk_data to pk_data
  joined_data <- merge(pk_data, ordered_pk_data, by = "PARAMCD", suffixes = c("", ".y"))

  joined_data <- joined_data[, -grep(".*.y$", colnames(joined_data))]

  joined_data$TLG_ORDER <- as.numeric(joined_data$TLG_ORDER)

  # Then order PARAM based on this column
  joined_data$PARAM <- factor(joined_data$PARAM,
    levels = unique(joined_data$PARAM[order(joined_data$TLG_ORDER)]),
    ordered = TRUE
  )

  joined_data$TLG_DISPLAY <- factor(joined_data$TLG_DISPLAY,
    levels = unique(joined_data$TLG_DISPLAY[order(joined_data$TLG_ORDER)]),
    ordered = TRUE
  )

  joined_data
}

#' Summarize change from baseline values or absolute baseline values
#'
#' @description `r lifecycle::badge("stable")`
#'
#' The analyze function [summarize_change()] creates a layout element to summarize the change from baseline or absolute
#' baseline values. The primary analysis variable `vars` indicates the numerical change from baseline results.
#'
#' Required secondary analysis variables `value` and `baseline_flag` can be supplied to the function via
#' the `variables` argument. The `value` element should be the name of the analysis value variable, and the
#' `baseline_flag` element should be the name of the flag variable that indicates whether or not records contain
#' baseline values. Depending on the baseline flag given, either the absolute baseline values (at baseline)
#' or the change from baseline values (post-baseline) are then summarized.
#'
#' @inheritParams argument_convention
#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("analyze_vars_numeric)`
#'   to see available statistics for this function.
#'
#' @name summarize_change
#' @order 1
NULL

#' @describeIn summarize_change Statistics function that summarizes baseline or post-baseline visits.
#'
#' @return
#' * `s_change_from_baseline()` returns the same values returned by [s_summary.numeric()].
#'
#' @note The data in `df` must be either all be from baseline or post-baseline visits. Otherwise
#'   an error will be thrown.
#'
#' @keywords internal
s_change_from_baseline <- function(df,
                                   .var,
                                   variables,
                                   na.rm = TRUE, # nolint
                                   ...) {
  checkmate::assert_numeric(df[[variables$value]])
  checkmate::assert_numeric(df[[.var]])
  checkmate::assert_logical(df[[variables$baseline_flag]])
  checkmate::assert_vector(unique(df[[variables$baseline_flag]]), max.len = 1)
  assert_df_with_variables(df, c(variables, list(chg = .var)))

  combined <- ifelse(
    df[[variables$baseline_flag]],
    df[[variables$value]],
    df[[.var]]
  )
  if (is.logical(combined) && identical(length(combined), 0L)) {
    combined <- numeric(0)
  }
  s_summary(combined, na.rm = na.rm, ...)
}

#' @describeIn summarize_change Formatted analysis function which is used as `afun` in `summarize_change()`.
#'
#' @return
#' * `a_change_from_baseline()` returns the corresponding list with formatted [rtables::CellValue()].
#'
#' @keywords internal
a_change_from_baseline <- make_afun(
  s_change_from_baseline,
  .formats = c(
    n = "xx",
    mean_sd = "xx.xx (xx.xx)",
    mean_se = "xx.xx (xx.xx)",
    median = "xx.xx",
    range = "xx.xx - xx.xx",
    mean_ci = "(xx.xx, xx.xx)",
    median_ci = "(xx.xx, xx.xx)",
    mean_pval = "xx.xx"
  ),
  .labels = c(
    mean_sd = "Mean (SD)",
    mean_se = "Mean (SE)",
    median = "Median",
    range = "Min - Max"
  )
)

#' @describeIn summarize_change Layout-creating function which can take statistics function arguments
#'   and additional format arguments. This function is a wrapper for [rtables::analyze()].
#'
#' @return
#' * `summarize_change()` returns a layout object suitable for passing to further layouting functions,
#'   or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
#'   the statistics from `s_change_from_baseline()` to the table layout.
#'
#' @note To be used after a split on visits in the layout, such that each data subset only contains
#'   either baseline or post-baseline data.
#'
#' @examples
#' library(dplyr)
#'
#' ## Fabricate dataset
#' dta_test <- data.frame(
#'   USUBJID = rep(1:6, each = 3),
#'   AVISIT = rep(paste0("V", 1:3), 6),
#'   ARM = rep(LETTERS[1:3], rep(6, 3)),
#'   AVAL = c(9:1, rep(NA, 9))
#' ) %>%
#'   mutate(ABLFLL = AVISIT == "V1") %>%
#'   group_by(USUBJID) %>%
#'   mutate(
#'     BLVAL = AVAL[ABLFLL],
#'     CHG = AVAL - BLVAL
#'   ) %>%
#'   ungroup()
#'
#' results <- basic_table() %>%
#'   split_cols_by("ARM") %>%
#'   split_rows_by("AVISIT") %>%
#'   summarize_change("CHG", variables = list(value = "AVAL", baseline_flag = "ABLFLL")) %>%
#'   build_table(dta_test)
#'
#' results
#'
#' @export
#' @order 2
summarize_change <- function(lyt,
                             vars,
                             variables,
                             na_str = default_na_str(),
                             nested = TRUE,
                             ...,
                             table_names = vars,
                             .stats = c("n", "mean_sd", "median", "range"),
                             .formats = NULL,
                             .labels = NULL,
                             .indent_mods = NULL) {
  extra_args <- list(variables = variables, ...)

  afun <- make_afun(
    a_change_from_baseline,
    .stats = .stats,
    .formats = .formats,
    .labels = .labels,
    .indent_mods = .indent_mods
  )

  analyze(
    lyt,
    vars,
    afun = afun,
    na_str = na_str,
    nested = nested,
    extra_args = extra_args,
    table_names = table_names
  )
}

#' Additional assertions to use with `checkmate`
#'
#' Additional assertion functions which can be used together with the `checkmate` package.
#'
#' @inheritParams checkmate::assert_factor
#' @param x (`any`)\cr object to test.
#' @param df (`data.frame`)\cr data set to test.
#' @param variables (named `list` of `character`)\cr list of variables to test.
#' @param include_boundaries (`flag`)\cr whether to include boundaries when testing
#'   for proportions.
#' @param na_level (`string`)\cr the string you have been using to represent NA or
#'   missing data. For `NA` values please consider using directly [is.na()] or
#'   similar approaches.
#'
#' @return Nothing if assertion passes, otherwise prints the error message.
#'
#' @name assertions
NULL

check_list_of_variables <- function(x) {
  # drop NULL elements in list
  x <- Filter(Negate(is.null), x)

  res <- checkmate::check_list(x,
    names = "named",
    min.len = 1,
    any.missing = FALSE,
    types = "character"
  )
  # no empty strings allowed
  if (isTRUE(res)) {
    res <- checkmate::check_character(unlist(x), min.chars = 1)
  }
  return(res)
}
#' @describeIn assertions Checks whether `x` is a valid list of variable names.
#'   `NULL` elements of the list `x` are dropped with `Filter(Negate(is.null), x)`.
#'
#' @keywords internal
assert_list_of_variables <- checkmate::makeAssertionFunction(check_list_of_variables)

check_df_with_variables <- function(df, variables, na_level = NULL) {
  checkmate::assert_data_frame(df)
  assert_list_of_variables(variables)

  # flag for equal variables and column names
  err_flag <- all(unlist(variables) %in% colnames(df))
  checkmate::assert_flag(err_flag)

  if (isFALSE(err_flag)) {
    vars <- setdiff(unlist(variables), colnames(df))
    return(paste(
      deparse(substitute(df)),
      "does not contain all specified variables as column names. Missing from data frame:",
      paste(vars, collapse = ", ")
    ))
  }
  # checking if na_level is present and in which column
  if (!is.null(na_level)) {
    checkmate::assert_string(na_level)
    res <- unlist(lapply(as.list(df)[unlist(variables)], function(x) any(x == na_level)))
    if (any(res)) {
      return(paste0(
        deparse(substitute(df)), " contains explicit na_level (", na_level,
        ") in the following columns: ", paste0(unlist(variables)[res],
          collapse = ", "
        )
      ))
    }
  }
  return(TRUE)
}
#' @describeIn assertions Check whether `df` is a data frame with the analysis `variables`.
#'   Please notice how this produces an error when not all variables are present in the
#'   data.frame while the opposite is not required.
#'
#' @keywords internal
assert_df_with_variables <- checkmate::makeAssertionFunction(check_df_with_variables)

check_valid_factor <- function(x,
                               min.levels = 1, # nolint
                               max.levels = NULL, # nolint
                               null.ok = TRUE, # nolint
                               any.missing = TRUE, # nolint
                               n.levels = NULL, # nolint
                               len = NULL) {
  # checks on levels insertion
  checkmate::assert_int(min.levels, lower = 1)

  # main factor check
  res <- checkmate::check_factor(x,
    min.levels = min.levels,
    null.ok = null.ok,
    max.levels = max.levels,
    any.missing = any.missing,
    n.levels = n.levels
  )

  # no empty strings allowed
  if (isTRUE(res)) {
    res <- checkmate::check_character(levels(x), min.chars = 1)
  }

  return(res)
}
#' @describeIn assertions Check whether `x` is a valid factor (i.e. has levels and no empty
#'   string levels). Note that `NULL` and `NA` elements are allowed.
#'
#' @keywords internal
assert_valid_factor <- checkmate::makeAssertionFunction(check_valid_factor)

check_df_with_factors <- function(df,
                                  variables,
                                  min.levels = 1, # nolint
                                  max.levels = NULL, # nolint
                                  any.missing = TRUE, # nolint
                                  na_level = NULL) {
  res <- check_df_with_variables(df, variables, na_level)
  # checking if all the columns specified by variables are valid factors
  if (isTRUE(res)) {
    # searching the data.frame with selected columns (variables) as a list
    res <- lapply(
      X = as.list(df)[unlist(variables)],
      FUN = check_valid_factor,
      min.levels = min.levels,
      max.levels = max.levels,
      any.missing = any.missing
    )
    res_lo <- unlist(vapply(res, Negate(isTRUE), logical(1)))
    if (any(res_lo)) {
      return(paste0(
        deparse(substitute(df)), " does not contain only factor variables among:",
        "\n* Column `", paste0(unlist(variables)[res_lo],
          "` of the data.frame -> ", res[res_lo],
          collapse = "\n* "
        )
      ))
    } else {
      res <- TRUE
    }
  }
  return(res)
}

#' @describeIn assertions Check whether `df` is a data frame where the analysis `variables`
#'   are all factors. Note that the creation of `NA` by direct call of `factor()` will
#'   trim `NA` levels out of the vector list itself.
#'
#' @keywords internal
assert_df_with_factors <- checkmate::makeAssertionFunction(check_df_with_factors)

#' @describeIn assertions Check whether `x` is a proportion: number between 0 and 1.
#'
#' @keywords internal
assert_proportion_value <- function(x, include_boundaries = FALSE) {
  checkmate::assert_number(x, lower = 0, upper = 1)
  checkmate::assert_flag(include_boundaries)
  if (isFALSE(include_boundaries)) {
    checkmate::assert_true(x > 0)
    checkmate::assert_true(x < 1)
  }
}

#' Horizontal waterfall plot
#'
#' @description `r lifecycle::badge("stable")`
#'
#' This basic waterfall plot visualizes a quantity `height` ordered by value with some markup.
#'
#' @param height (`numeric`)\cr vector containing values to be plotted as the waterfall bars.
#' @param id (`character`)\cr vector containing identifiers to use as the x-axis label for the waterfall bars.
#' @param col (`character`)\cr color(s).
#' @param col_var (`factor`, `character`, or `NULL`)\cr categorical variable for bar coloring. `NULL` by default.
#' @param xlab (`string`)\cr x label. Default is `"ID"`.
#' @param ylab (`string`)\cr y label. Default is `"Value"`.
#' @param title (`string`)\cr text to be displayed as plot title.
#' @param col_legend_title (`string`)\cr text to be displayed as legend title.
#'
#' @return A `ggplot` waterfall plot.
#'
#' @examples
#' library(dplyr)
#' library(nestcolor)
#'
#' g_waterfall(height = c(3, 5, -1), id = letters[1:3])
#'
#' g_waterfall(
#'   height = c(3, 5, -1),
#'   id = letters[1:3],
#'   col_var = letters[1:3]
#' )
#'
#' adsl_f <- tern_ex_adsl %>%
#'   select(USUBJID, STUDYID, ARM, ARMCD, SEX)
#'
#' adrs_f <- tern_ex_adrs %>%
#'   filter(PARAMCD == "OVRINV") %>%
#'   mutate(pchg = rnorm(n(), 10, 50))
#'
#' adrs_f <- head(adrs_f, 30)
#' adrs_f <- adrs_f[!duplicated(adrs_f$USUBJID), ]
#' head(adrs_f)
#'
#' g_waterfall(
#'   height = adrs_f$pchg,
#'   id = adrs_f$USUBJID,
#'   col_var = adrs_f$AVALC
#' )
#'
#' g_waterfall(
#'   height = adrs_f$pchg,
#'   id = paste("asdfdsfdsfsd", adrs_f$USUBJID),
#'   col_var = adrs_f$SEX
#' )
#'
#' g_waterfall(
#'   height = adrs_f$pchg,
#'   id = paste("asdfdsfdsfsd", adrs_f$USUBJID),
#'   xlab = "ID",
#'   ylab = "Percentage Change",
#'   title = "Waterfall plot"
#' )
#'
#' @export
g_waterfall <- function(height,
                        id,
                        col_var = NULL,
                        col = getOption("ggplot2.discrete.colour"),
                        xlab = NULL,
                        ylab = NULL,
                        col_legend_title = NULL,
                        title = NULL) {
  if (!is.null(col_var)) {
    check_same_n(height = height, id = id, col_var = col_var)
  } else {
    check_same_n(height = height, id = id)
  }

  checkmate::assert_multi_class(col_var, c("character", "factor"), null.ok = TRUE)
  checkmate::assert_character(col, null.ok = TRUE)

  xlabel <- deparse(substitute(id))
  ylabel <- deparse(substitute(height))

  col_label <- if (!missing(col_var)) {
    deparse(substitute(col_var))
  }

  xlab <- if (is.null(xlab)) xlabel else xlab
  ylab <- if (is.null(ylab)) ylabel else ylab
  col_legend_title <- if (is.null(col_legend_title)) col_label else col_legend_title

  plot_data <- data.frame(
    height = height,
    id = as.character(id),
    col_var = if (is.null(col_var)) "x" else to_n(col_var, length(height)),
    stringsAsFactors = FALSE
  )

  plot_data_ord <- plot_data[order(plot_data$height, decreasing = TRUE), ]

  p <- ggplot2::ggplot(plot_data_ord, ggplot2::aes(x = factor(id, levels = id), y = height)) +
    ggplot2::geom_col() +
    ggplot2::geom_text(
      label = format(plot_data_ord$height, digits = 2),
      vjust = ifelse(plot_data_ord$height >= 0, -0.5, 1.5)
    ) +
    ggplot2::xlab(xlab) +
    ggplot2::ylab(ylab) +
    ggplot2::theme(axis.text.x = ggplot2::element_text(angle = 90, hjust = 0, vjust = .5))

  if (!is.null(col_var)) {
    p <- p +
      ggplot2::aes(fill = col_var) +
      ggplot2::labs(fill = col_legend_title) +
      ggplot2::theme(
        legend.position = "bottom",
        legend.background = ggplot2::element_blank(),
        legend.title = ggplot2::element_text(face = "bold"),
        legend.box.background = ggplot2::element_rect(colour = "black")
      )
  }

  if (!is.null(col)) {
    p <- p +
      ggplot2::scale_fill_manual(values = col)
  }

  if (!is.null(title)) {
    p <- p +
      ggplot2::labs(title = title) +
      ggplot2::theme(plot.title = ggplot2::element_text(face = "bold"))
  }

  p
}

#' Cox regression helper function for interactions
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Test and estimate the effect of a treatment in interaction with a covariate.
#' The effect is estimated as the HR of the tested treatment for a given level
#' of the covariate, in comparison to the treatment control.
#'
#' @inheritParams argument_convention
#' @param x (`numeric` or `factor`)\cr the values of the covariate to be tested.
#' @param effect (`string`)\cr the name of the effect to be tested and estimated.
#' @param covar (`string`)\cr the name of the covariate in the model.
#' @param mod (`coxph`)\cr the Cox regression model.
#' @param label (`string`)\cr the label to be returned as `term_label`.
#' @param control (`list`)\cr a list of controls as returned by [control_coxreg()].
#' @param ... see methods.
#'
#' @examples
#' library(survival)
#'
#' set.seed(1, kind = "Mersenne-Twister")
#'
#' # Testing dataset [survival::bladder].
#' dta_bladder <- with(
#'   data = bladder[bladder$enum < 5, ],
#'   data.frame(
#'     time = stop,
#'     status = event,
#'     armcd = as.factor(rx),
#'     covar1 = as.factor(enum),
#'     covar2 = factor(
#'       sample(as.factor(enum)),
#'       levels = 1:4,
#'       labels = c("F", "F", "M", "M")
#'     )
#'   )
#' )
#' labels <- c("armcd" = "ARM", "covar1" = "A Covariate Label", "covar2" = "Sex (F/M)")
#' formatters::var_labels(dta_bladder)[names(labels)] <- labels
#' dta_bladder$age <- sample(20:60, size = nrow(dta_bladder), replace = TRUE)
#'
#' plot(
#'   survfit(Surv(time, status) ~ armcd + covar1, data = dta_bladder),
#'   lty = 2:4,
#'   xlab = "Months",
#'   col = c("blue1", "blue2", "blue3", "blue4", "red1", "red2", "red3", "red4")
#' )
#'
#' @name cox_regression_inter
NULL

#' @describeIn cox_regression_inter S3 generic helper function to determine interaction effect.
#'
#' @return
#' * `h_coxreg_inter_effect()` returns a `data.frame` of covariate interaction effects consisting of the following
#'   variables: `effect`, `term`, `term_label`, `level`, `n`, `hr`, `lcl`, `ucl`, `pval`, and `pval_inter`.
#'
#' @export
h_coxreg_inter_effect <- function(x,
                                  effect,
                                  covar,
                                  mod,
                                  label,
                                  control,
                                  ...) {
  UseMethod("h_coxreg_inter_effect", x)
}

#' @describeIn cox_regression_inter Method for `numeric` class. Estimates the interaction with a `numeric` covariate.
#'
#' @method h_coxreg_inter_effect numeric
#'
#' @param at (`list`)\cr a list with items named after the covariate, every
#'   item is a vector of levels at which the interaction should be estimated.
#'
#' @export
h_coxreg_inter_effect.numeric <- function(x,
                                          effect,
                                          covar,
                                          mod,
                                          label,
                                          control,
                                          at,
                                          ...) {
  betas <- stats::coef(mod)
  attrs <- attr(stats::terms(mod), "term.labels")
  term_indices <- grep(
    pattern = effect,
    x = attrs[!grepl("strata\\(", attrs)]
  )
  checkmate::assert_vector(term_indices, len = 2)
  betas <- betas[term_indices]
  betas_var <- diag(stats::vcov(mod))[term_indices]
  betas_cov <- stats::vcov(mod)[term_indices[1], term_indices[2]]
  xval <- if (is.null(at[[covar]])) {
    stats::median(x)
  } else {
    at[[covar]]
  }
  effect_index <- !grepl(covar, names(betas))
  coef_hat <- betas[effect_index] + xval * betas[!effect_index]
  coef_se <- sqrt(
    betas_var[effect_index] +
      xval ^ 2 * betas_var[!effect_index] + # styler: off
      2 * xval * betas_cov
  )
  q_norm <- stats::qnorm((1 + control$conf_level) / 2)
  data.frame(
    effect = "Covariate:",
    term = rep(covar, length(xval)),
    term_label = paste0("  ", xval),
    level = as.character(xval),
    n = NA,
    hr = exp(coef_hat),
    lcl = exp(coef_hat - q_norm * coef_se),
    ucl = exp(coef_hat + q_norm * coef_se),
    pval = NA,
    pval_inter = NA,
    stringsAsFactors = FALSE
  )
}

#' @describeIn cox_regression_inter Method for `factor` class. Estimate the interaction with a `factor` covariate.
#'
#' @method h_coxreg_inter_effect factor
#'
#' @param data (`data.frame`)\cr the data frame on which the model was fit.
#'
#' @export
h_coxreg_inter_effect.factor <- function(x,
                                         effect,
                                         covar,
                                         mod,
                                         label,
                                         control,
                                         data,
                                         ...) {
  lvl_given <- levels(x)
  y <- h_coxreg_inter_estimations(
    variable = effect, given = covar,
    lvl_var = levels(data[[effect]]),
    lvl_given = lvl_given,
    mod = mod,
    conf_level = 0.95
  )[[1]]

  data.frame(
    effect = "Covariate:",
    term = rep(covar, nrow(y)),
    term_label = paste0("  ", lvl_given),
    level = lvl_given,
    n = NA,
    hr = y[, "hr"],
    lcl = y[, "lcl"],
    ucl = y[, "ucl"],
    pval = NA,
    pval_inter = NA,
    stringsAsFactors = FALSE
  )
}

#' @describeIn cox_regression_inter Method for `character` class. Estimate the interaction with a `character` covariate.
#'   This makes an automatic conversion to `factor` and then forwards to the method for factors.
#'
#' @method h_coxreg_inter_effect character
#'
#' @note
#' * Automatic conversion of character to factor does not guarantee results can be generated correctly. It is
#'   therefore better to always pre-process the dataset such that factors are manually created from character
#'   variables before passing the dataset to [rtables::build_table()].
#'
#' @export
h_coxreg_inter_effect.character <- function(x,
                                            effect,
                                            covar,
                                            mod,
                                            label,
                                            control,
                                            data,
                                            ...) {
  y <- as.factor(x)

  h_coxreg_inter_effect(
    x = y,
    effect = effect,
    covar = covar,
    mod = mod,
    label = label,
    control = control,
    data = data,
    ...
  )
}

#' @describeIn cox_regression_inter A higher level function to get
#'   the results of the interaction test and the estimated values.
#'
#' @return
#' * `h_coxreg_extract_interaction()` returns the result of an interaction test and the estimated values. If
#'   no interaction, [h_coxreg_univar_extract()] is applied instead.
#'
#' @examples
#' mod <- coxph(Surv(time, status) ~ armcd * covar1, data = dta_bladder)
#' h_coxreg_extract_interaction(
#'   mod = mod, effect = "armcd", covar = "covar1", data = dta_bladder,
#'   control = control_coxreg()
#' )
#'
#' @export
h_coxreg_extract_interaction <- function(effect,
                                         covar,
                                         mod,
                                         data,
                                         at,
                                         control) {
  if (!any(attr(stats::terms(mod), "order") == 2)) {
    y <- h_coxreg_univar_extract(
      effect = effect, covar = covar, mod = mod, data = data, control = control
    )
    y$pval_inter <- NA
    y
  } else {
    test_statistic <- c(wald = "Wald", likelihood = "LR")[control$pval_method]

    # Test the main treatment effect.
    mod_aov <- muffled_car_anova(mod, test_statistic)
    sum_anova <- broom::tidy(mod_aov)
    pval <- sum_anova[sum_anova$term == effect, ][["p.value"]]

    # Test the interaction effect.
    pval_inter <- sum_anova[grep(":", sum_anova$term), ][["p.value"]]
    covar_test <- data.frame(
      effect = "Covariate:",
      term = covar,
      term_label = unname(labels_or_names(data[covar])),
      level = "",
      n = mod$n, hr = NA, lcl = NA, ucl = NA, pval = pval,
      pval_inter = pval_inter,
      stringsAsFactors = FALSE
    )
    # Estimate the interaction.
    y <- h_coxreg_inter_effect(
      data[[covar]],
      covar = covar,
      effect = effect,
      mod = mod,
      label = unname(labels_or_names(data[covar])),
      at = at,
      control = control,
      data = data
    )
    rbind(covar_test, y)
  }
}

#' @describeIn cox_regression_inter Hazard ratio estimation in interactions.
#'
#' @param variable,given (`string`)\cr the name of variables in interaction. We seek the estimation
#'   of the levels of `variable` given the levels of `given`.
#' @param lvl_var,lvl_given (`character`)\cr corresponding levels as given by [levels()].
#' @param mod (`coxph`)\cr a fitted Cox regression model (see [survival::coxph()]).
#'
#' @details Given the cox regression investigating the effect of Arm (A, B, C; reference A)
#'   and Sex (F, M; reference Female) and the model being abbreviated: y ~ Arm + Sex + Arm:Sex.
#'   The cox regression estimates the coefficients along with a variance-covariance matrix for:
#'
#'   - b1 (arm b), b2 (arm c)
#'   - b3 (sex m)
#'   - b4 (arm b: sex m), b5 (arm c: sex m)
#'
#'   The estimation of the Hazard Ratio for arm C/sex M is given in reference
#'   to arm A/Sex M by exp(b2 + b3 + b5)/ exp(b3) = exp(b2 + b5).
#'   The interaction coefficient is deduced by b2 + b5 while the standard error
#'   is obtained as $sqrt(Var b2 + Var b5 + 2 * covariance (b2,b5))$.
#'
#' @return
#' * `h_coxreg_inter_estimations()` returns a list of matrices (one per level of variable) with rows corresponding
#'   to the combinations of `variable` and `given`, with columns:
#'   * `coef_hat`: Estimation of the coefficient.
#'   * `coef_se`: Standard error of the estimation.
#'   * `hr`: Hazard ratio.
#'   * `lcl, ucl`: Lower/upper confidence limit of the hazard ratio.
#'
#' @examples
#' mod <- coxph(Surv(time, status) ~ armcd * covar1, data = dta_bladder)
#' result <- h_coxreg_inter_estimations(
#'   variable = "armcd", given = "covar1",
#'   lvl_var = levels(dta_bladder$armcd),
#'   lvl_given = levels(dta_bladder$covar1),
#'   mod = mod, conf_level = .95
#' )
#' result
#'
#' @export
h_coxreg_inter_estimations <- function(variable,
                                       given,
                                       lvl_var,
                                       lvl_given,
                                       mod,
                                       conf_level = 0.95) {
  var_lvl <- paste0(variable, lvl_var[-1]) # [-1]: reference level
  giv_lvl <- paste0(given, lvl_given)
  design_mat <- expand.grid(variable = var_lvl, given = giv_lvl)
  design_mat <- design_mat[order(design_mat$variable, design_mat$given), ]
  design_mat <- within(
    data = design_mat,
    expr = {
      inter <- paste0(variable, ":", given)
      rev_inter <- paste0(given, ":", variable)
    }
  )
  split_by_variable <- design_mat$variable
  interaction_names <- paste(design_mat$variable, design_mat$given, sep = "/")

  mmat <- stats::model.matrix(mod)[1, ]
  mmat[!mmat == 0] <- 0

  design_mat <- apply(
    X = design_mat, MARGIN = 1, FUN = function(x) {
      mmat[names(mmat) %in% x[-which(names(x) == "given")]] <- 1
      mmat
    }
  )
  colnames(design_mat) <- interaction_names

  coef <- stats::coef(mod)
  vcov <- stats::vcov(mod)
  betas <- as.matrix(coef)
  coef_hat <- t(design_mat) %*% betas
  dimnames(coef_hat)[2] <- "coef"
  coef_se <- apply(
    design_mat, 2,
    function(x) {
      vcov_el <- as.logical(x)
      y <- vcov[vcov_el, vcov_el]
      y <- sum(y)
      y <- sqrt(y)
      return(y)
    }
  )
  q_norm <- stats::qnorm((1 + conf_level) / 2)
  y <- cbind(coef_hat, `se(coef)` = coef_se)
  y <- apply(y, 1, function(x) {
    x["hr"] <- exp(x["coef"])
    x["lcl"] <- exp(x["coef"] - q_norm * x["se(coef)"])
    x["ucl"] <- exp(x["coef"] + q_norm * x["se(coef)"])
    x
  })
  y <- t(y)
  y <- by(y, split_by_variable, identity)
  y <- lapply(y, as.matrix)
  attr(y, "details") <- paste0(
    "Estimations of ", variable,
    " hazard ratio given the level of ", given, " compared to ",
    variable, " level ", lvl_var[1], "."
  )
  y
}

#' Helper functions for tabulating biomarker effects on binary response by subgroup
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Helper functions which are documented here separately to not confuse the user
#' when reading about the user-facing functions.
#'
#' @inheritParams response_biomarkers_subgroups
#' @inheritParams extract_rsp_biomarkers
#' @inheritParams argument_convention
#'
#' @examples
#' library(dplyr)
#' library(forcats)
#'
#' adrs <- tern_ex_adrs
#' adrs_labels <- formatters::var_labels(adrs)
#'
#' adrs_f <- adrs %>%
#'   filter(PARAMCD == "BESRSPI") %>%
#'   mutate(rsp = AVALC == "CR")
#' formatters::var_labels(adrs_f) <- c(adrs_labels, "Response")
#'
#' @name h_response_biomarkers_subgroups
NULL

#' @describeIn h_response_biomarkers_subgroups helps with converting the "response" function variable list
#'   to the "logistic regression" variable list. The reason is that currently there is an
#'   inconsistency between the variable names accepted by `extract_rsp_subgroups()` and `fit_logistic()`.
#'
#' @param biomarker (`string`)\cr the name of the biomarker variable.
#'
#' @return
#' * `h_rsp_to_logistic_variables()` returns a named `list` of elements `response`, `arm`, `covariates`, and `strata`.
#'
#' @examples
#' # This is how the variable list is converted internally.
#' h_rsp_to_logistic_variables(
#'   variables = list(
#'     rsp = "RSP",
#'     covariates = c("A", "B"),
#'     strata = "D"
#'   ),
#'   biomarker = "AGE"
#' )
#'
#' @export
h_rsp_to_logistic_variables <- function(variables, biomarker) {
  if ("strat" %in% names(variables)) {
    warning(
      "Warning: the `strat` element name of the `variables` list argument to `h_rsp_to_logistic_variables() ",
      "was deprecated in tern 0.9.4.\n  ",
      "Please use the name `strata` instead of `strat` in the `variables` argument."
    )
    variables[["strata"]] <- variables[["strat"]]
  }
  checkmate::assert_list(variables)
  checkmate::assert_string(variables$rsp)
  checkmate::assert_string(biomarker)
  list(
    response = variables$rsp,
    arm = biomarker,
    covariates = variables$covariates,
    strata = variables$strata
  )
}

#' @describeIn h_response_biomarkers_subgroups prepares estimates for number of responses, patients and
#'   overall response rate, as well as odds ratio estimates, confidence intervals and p-values, for multiple
#'   biomarkers in a given single data set.
#'   `variables` corresponds to names of variables found in `data`, passed as a named list and requires elements
#'   `rsp` and `biomarkers` (vector of continuous biomarker variables) and optionally `covariates`
#'   and `strata`.
#'
#' @return
#' * `h_logistic_mult_cont_df()` returns a `data.frame` containing estimates and statistics for the selected biomarkers.
#'
#' @examples
#' # For a single population, estimate separately the effects
#' # of two biomarkers.
#' df <- h_logistic_mult_cont_df(
#'   variables = list(
#'     rsp = "rsp",
#'     biomarkers = c("BMRKR1", "AGE"),
#'     covariates = "SEX"
#'   ),
#'   data = adrs_f
#' )
#' df
#'
#' # If the data set is empty, still the corresponding rows with missings are returned.
#' h_coxreg_mult_cont_df(
#'   variables = list(
#'     rsp = "rsp",
#'     biomarkers = c("BMRKR1", "AGE"),
#'     covariates = "SEX",
#'     strata = "STRATA1"
#'   ),
#'   data = adrs_f[NULL, ]
#' )
#'
#' @export
h_logistic_mult_cont_df <- function(variables,
                                    data,
                                    control = control_logistic()) {
  if ("strat" %in% names(variables)) {
    warning(
      "Warning: the `strat` element name of the `variables` list argument to `h_logistic_mult_cont_df() ",
      "was deprecated in tern 0.9.4.\n  ",
      "Please use the name `strata` instead of `strat` in the `variables` argument."
    )
    variables[["strata"]] <- variables[["strat"]]
  }
  assert_df_with_variables(data, variables)

  checkmate::assert_character(variables$biomarkers, min.len = 1, any.missing = FALSE)
  checkmate::assert_list(control, names = "named")

  conf_level <- control[["conf_level"]]
  pval_label <- "p-value (Wald)"

  # If there is any data, run model, otherwise return empty results.
  if (nrow(data) > 0) {
    bm_cols <- match(variables$biomarkers, names(data))
    l_result <- lapply(variables$biomarkers, function(bm) {
      model_fit <- fit_logistic(
        variables = h_rsp_to_logistic_variables(variables, bm),
        data = data,
        response_definition = control$response_definition
      )
      result <- h_logistic_simple_terms(
        x = bm,
        fit_glm = model_fit,
        conf_level = control$conf_level
      )
      resp_vector <- if (inherits(model_fit, "glm")) {
        model_fit$model[[variables$rsp]]
      } else {
        as.logical(as.matrix(model_fit$y)[, "status"])
      }
      data.frame(
        # Dummy column needed downstream to create a nested header.
        biomarker = bm,
        biomarker_label = formatters::var_labels(data[bm], fill = TRUE),
        n_tot = length(resp_vector),
        n_rsp = sum(resp_vector),
        prop = mean(resp_vector),
        or = as.numeric(result[1L, "odds_ratio"]),
        lcl = as.numeric(result[1L, "lcl"]),
        ucl = as.numeric(result[1L, "ucl"]),
        conf_level = conf_level,
        pval = as.numeric(result[1L, "pvalue"]),
        pval_label = pval_label,
        stringsAsFactors = FALSE
      )
    })
    do.call(rbind, args = c(l_result, make.row.names = FALSE))
  } else {
    data.frame(
      biomarker = variables$biomarkers,
      biomarker_label = formatters::var_labels(data[variables$biomarkers], fill = TRUE),
      n_tot = 0L,
      n_rsp = 0L,
      prop = NA,
      or = NA,
      lcl = NA,
      ucl = NA,
      conf_level = conf_level,
      pval = NA,
      pval_label = pval_label,
      row.names = seq_along(variables$biomarkers),
      stringsAsFactors = FALSE
    )
  }
}

#' @describeIn h_response_biomarkers_subgroups Prepares a single sub-table given a `df_sub` containing
#'   the results for a single biomarker.
#'
#' @param df (`data.frame`)\cr results for a single biomarker, as part of what is
#'   returned by [extract_rsp_biomarkers()] (it needs a couple of columns which are
#'   added by that high-level function relative to what is returned by [h_logistic_mult_cont_df()],
#'   see the example).
#'
#' @return
#' * `h_tab_rsp_one_biomarker()` returns an `rtables` table object with the given statistics arranged in columns.
#'
#' @examples
#' # Starting from above `df`, zoom in on one biomarker and add required columns.
#' df1 <- df[1, ]
#' df1$subgroup <- "All patients"
#' df1$row_type <- "content"
#' df1$var <- "ALL"
#' df1$var_label <- "All patients"
#'
#' h_tab_rsp_one_biomarker(
#'   df1,
#'   vars = c("n_tot", "n_rsp", "prop", "or", "ci", "pval")
#' )
#'
#' @export
h_tab_rsp_one_biomarker <- function(df,
                                    vars,
                                    na_str = default_na_str(),
                                    .indent_mods = 0L) {
  afuns <- a_response_subgroups(na_str = na_str)[vars]
  colvars <- d_rsp_subgroups_colvars(
    vars,
    conf_level = df$conf_level[1],
    method = df$pval_label[1]
  )
  h_tab_one_biomarker(
    df = df,
    afuns = afuns,
    colvars = colvars,
    na_str = na_str,
    .indent_mods = .indent_mods
  )
}

#' Class for `CombinationFunction`
#'
#' @description `r lifecycle::badge("stable")`
#'
#' `CombinationFunction` is an S4 class which extends standard functions. These are special functions that
#' can be combined and negated with the logical operators.
#'
#' @param e1 (`CombinationFunction`)\cr left hand side of logical operator.
#' @param e2 (`CombinationFunction`)\cr right hand side of logical operator.
#' @param x (`CombinationFunction`)\cr the function which should be negated.
#'
#' @return A logical value indicating whether the left hand side of the equation equals the right hand side.
#'
#' @examples
#' higher <- function(a) {
#'   force(a)
#'   CombinationFunction(
#'     function(x) {
#'       x > a
#'     }
#'   )
#' }
#'
#' lower <- function(b) {
#'   force(b)
#'   CombinationFunction(
#'     function(x) {
#'       x < b
#'     }
#'   )
#' }
#'
#' c1 <- higher(5)
#' c2 <- lower(10)
#' c3 <- higher(5) & lower(10)
#' c3(7)
#'
#' @name combination_function
#' @aliases CombinationFunction-class
#' @exportClass CombinationFunction
#' @export CombinationFunction
CombinationFunction <- methods::setClass("CombinationFunction", contains = "function") # nolint

#' @describeIn combination_function Logical "AND" combination of `CombinationFunction` functions.
#'   The resulting object is of the same class, and evaluates the two argument functions. The result
#'   is then the "AND" of the two individual results.
#'
#' @export
methods::setMethod(
  "&",
  signature = c(e1 = "CombinationFunction", e2 = "CombinationFunction"),
  definition = function(e1, e2) {
    CombinationFunction(function(...) {
      e1(...) && e2(...)
    })
  }
)

#' @describeIn combination_function Logical "OR" combination of `CombinationFunction` functions.
#'   The resulting object is of the same class, and evaluates the two argument functions. The result
#'   is then the "OR" of the two individual results.
#'
#' @export
methods::setMethod(
  "|",
  signature = c(e1 = "CombinationFunction", e2 = "CombinationFunction"),
  definition = function(e1, e2) {
    CombinationFunction(function(...) {
      e1(...) || e2(...)
    })
  }
)

#' @describeIn combination_function Logical negation of `CombinationFunction` functions.
#'   The resulting object is of the same class, and evaluates the original function. The result
#'   is then the opposite of this results.
#'
#' @export
methods::setMethod(
  "!",
  signature = c(x = "CombinationFunction"),
  definition = function(x) {
    CombinationFunction(function(...) {
      !x(...)
    })
  }
)

#' Summarize variables in columns
#'
#' @description `r lifecycle::badge("stable")`
#'
#' The analyze function [summarize_colvars()] uses the statistics function [s_summary()] to analyze variables that are
#' arranged in columns. The variables to analyze should be specified in the table layout via column splits (see
#' [rtables::split_cols_by()] and [rtables::split_cols_by_multivar()]) prior to using [summarize_colvars()].
#'
#' The function is a minimal wrapper for [rtables::analyze_colvars()], a function typically used to apply different
#' analysis methods in rows for each column variable. To use the analysis methods as column labels, please refer to
#' the [analyze_vars_in_cols()] function.
#'
#' @inheritParams argument_convention
#' @param ... arguments passed to [s_summary()].
#' @param .indent_mods (named `vector` of `integer`)\cr indent modifiers for the labels. Each element of the vector
#'   should be a name-value pair with name corresponding to a statistic specified in `.stats` and value the indentation
#'   for that statistic's row label.
#'
#' @return
#' A layout object suitable for passing to further layouting functions, or to [rtables::build_table()].
#' Adding this function to an `rtable` layout will summarize the given variables, arrange the output
#' in columns, and add it to the table layout.
#'
#' @seealso [rtables::split_cols_by_multivar()] and [`analyze_colvars_functions`].
#'
#' @examples
#' dta_test <- data.frame(
#'   USUBJID = rep(1:6, each = 3),
#'   PARAMCD = rep("lab", 6 * 3),
#'   AVISIT = rep(paste0("V", 1:3), 6),
#'   ARM = rep(LETTERS[1:3], rep(6, 3)),
#'   AVAL = c(9:1, rep(NA, 9)),
#'   CHG = c(1:9, rep(NA, 9))
#' )
#'
#' ## Default output within a `rtables` pipeline.
#' basic_table() %>%
#'   split_cols_by("ARM") %>%
#'   split_rows_by("AVISIT") %>%
#'   split_cols_by_multivar(vars = c("AVAL", "CHG")) %>%
#'   summarize_colvars() %>%
#'   build_table(dta_test)
#'
#' ## Selection of statistics, formats and labels also work.
#' basic_table() %>%
#'   split_cols_by("ARM") %>%
#'   split_rows_by("AVISIT") %>%
#'   split_cols_by_multivar(vars = c("AVAL", "CHG")) %>%
#'   summarize_colvars(
#'     .stats = c("n", "mean_sd"),
#'     .formats = c("mean_sd" = "xx.x, xx.x"),
#'     .labels = c(n = "n", mean_sd = "Mean, SD")
#'   ) %>%
#'   build_table(dta_test)
#'
#' ## Use arguments interpreted by `s_summary`.
#' basic_table() %>%
#'   split_cols_by("ARM") %>%
#'   split_rows_by("AVISIT") %>%
#'   split_cols_by_multivar(vars = c("AVAL", "CHG")) %>%
#'   summarize_colvars(na.rm = FALSE) %>%
#'   build_table(dta_test)
#'
#' @export
summarize_colvars <- function(lyt,
                              ...,
                              na_str = default_na_str(),
                              .stats = c("n", "mean_sd", "median", "range", "count_fraction"),
                              .formats = NULL,
                              .labels = NULL,
                              .indent_mods = NULL) {
  extra_args <- list(.stats = .stats, na_str = na_str, ...)
  if (!is.null(.formats)) extra_args[[".formats"]] <- .formats
  if (!is.null(.labels)) extra_args[[".labels"]] <- .labels
  if (!is.null(.indent_mods)) extra_args[[".indent_mods"]] <- .indent_mods

  analyze_colvars(
    lyt,
    afun = a_summary,
    na_str = na_str,
    extra_args = extra_args
  )
}

#' Apply 1/3 or 1/2 imputation rule to data
#'
#' @description `r lifecycle::badge("stable")`
#'
#' @inheritParams argument_convention
#' @param x_stats (named `list`)\cr a named list of statistics, typically the results of [s_summary()].
#' @param stat (`string`)\cr statistic to return the value/NA level of according to the imputation
#'   rule applied.
#' @param imp_rule (`string`)\cr imputation rule setting. Set to `"1/3"` to implement 1/3 imputation
#'   rule or `"1/2"` to implement 1/2 imputation rule.
#' @param post (`flag`)\cr whether the data corresponds to a post-dose time-point (defaults to `FALSE`).
#'   This parameter is only used when `imp_rule` is set to `"1/3"`.
#' @param avalcat_var (`string`)\cr name of variable that indicates whether a row in `df` corresponds
#'   to an analysis value in category `"BLQ"`, `"LTR"`, `"<PCLLOQ"`, or none of the above
#'   (defaults to `"AVALCAT1"`). Variable `avalcat_var` must be present in `df`.
#'
#' @return A `list` containing statistic value (`val`) and NA level (`na_str`) that should be displayed
#'   according to the specified imputation rule.
#'
#' @seealso [analyze_vars_in_cols()] where this function can be implemented by setting the `imp_rule`
#'   argument.
#'
#' @examples
#' set.seed(1)
#' df <- data.frame(
#'   AVAL = runif(50, 0, 1),
#'   AVALCAT1 = sample(c(1, "BLQ"), 50, replace = TRUE)
#' )
#' x_stats <- s_summary(df$AVAL)
#' imputation_rule(df, x_stats, "max", "1/3")
#' imputation_rule(df, x_stats, "geom_mean", "1/3")
#' imputation_rule(df, x_stats, "mean", "1/2")
#'
#' @export
imputation_rule <- function(df, x_stats, stat, imp_rule, post = FALSE, avalcat_var = "AVALCAT1") {
  checkmate::assert_choice(avalcat_var, names(df))
  checkmate::assert_choice(imp_rule, c("1/3", "1/2"))
  n_blq <- sum(grepl("BLQ|LTR|<[1-9]|<PCLLOQ", df[[avalcat_var]]))
  ltr_blq_ratio <- n_blq / max(1, nrow(df))

  # defaults
  val <- x_stats[[stat]]
  na_str <- "NE"

  if (imp_rule == "1/3") {
    if (!post && stat == "geom_mean") val <- NA # 1/3_pre_LT, 1/3_pre_GT
    if (ltr_blq_ratio > 1 / 3) {
      if (stat != "geom_mean") na_str <- "ND" # 1/3_pre_GT, 1/3_post_GT
      if (!post && !stat %in% c("median", "max")) val <- NA # 1/3_pre_GT
      if (post && !stat %in% c("median", "max", "geom_mean")) val <- NA # 1/3_post_GT
    }
  } else if (imp_rule == "1/2") {
    if (ltr_blq_ratio > 1 / 2 && !stat == "max") {
      val <- NA # 1/2_GT
      na_str <- "ND" # 1/2_GT
    }
  }

  list(val = val, na_str = na_str)
}

#' Analyze a pairwise Cox-PH model
#'
#' @description `r lifecycle::badge("stable")`
#'
#' The analyze function [coxph_pairwise()] creates a layout element to analyze a pairwise Cox-PH model.
#'
#' This function can return statistics including p-value, hazard ratio (HR), and HR confidence intervals from both
#' stratified and unstratified Cox-PH models. The variable(s) to be analyzed is specified via the `vars` argument and
#' any stratification factors via the `strata` argument.
#'
#' @inheritParams argument_convention
#' @inheritParams s_surv_time
#' @param strata (`character` or `NULL`)\cr variable names indicating stratification factors.
#' @param strat `r lifecycle::badge("deprecated")` Please use the `strata` argument instead.
#' @param control (`list`)\cr parameters for comparison details, specified by using the helper function
#'   [control_coxph()]. Some possible parameter options are:
#'   * `pval_method` (`string`)\cr p-value method for testing the null hypothesis that hazard ratio = 1. Default
#'     method is `"log-rank"` which comes from [survival::survdiff()], can also be set to `"wald"` or `"likelihood"`
#'     (from [survival::coxph()]).
#'   * `ties` (`string`)\cr specifying the method for tie handling. Default is `"efron"`,
#'     can also be set to `"breslow"` or `"exact"`. See more in [survival::coxph()].
#'   * `conf_level` (`proportion`)\cr confidence level of the interval for HR.
#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("coxph_pairwise")`
#'   to see available statistics for this function.
#'
#' @name survival_coxph_pairwise
#' @order 1
NULL

#' @describeIn survival_coxph_pairwise Statistics function which analyzes HR, CIs of HR, and p-value of a Cox-PH model.
#'
#' @return
#' * `s_coxph_pairwise()` returns the statistics:
#'   * `pvalue`: p-value to test the null hypothesis that hazard ratio = 1.
#'   * `hr`: Hazard ratio.
#'   * `hr_ci`: Confidence interval for hazard ratio.
#'   * `n_tot`: Total number of observations.
#'   * `n_tot_events`: Total number of events.
#'
#' @keywords internal
s_coxph_pairwise <- function(df,
                             .ref_group,
                             .in_ref_col,
                             .var,
                             is_event,
                             strata = NULL,
                             strat = lifecycle::deprecated(),
                             control = control_coxph()) {
  if (lifecycle::is_present(strat)) {
    lifecycle::deprecate_warn("0.9.4", "s_coxph_pairwise(strat)", "s_coxph_pairwise(strata)")
    strata <- strat
  }

  checkmate::assert_string(.var)
  checkmate::assert_numeric(df[[.var]])
  checkmate::assert_logical(df[[is_event]])
  assert_df_with_variables(df, list(tte = .var, is_event = is_event))
  pval_method <- control$pval_method
  ties <- control$ties
  conf_level <- control$conf_level

  if (.in_ref_col) {
    return(
      list(
        pvalue = formatters::with_label("", paste0("p-value (", pval_method, ")")),
        hr = formatters::with_label("", "Hazard Ratio"),
        hr_ci = formatters::with_label("", f_conf_level(conf_level)),
        n_tot = formatters::with_label("", "Total n"),
        n_tot_events = formatters::with_label("", "Total events")
      )
    )
  }
  data <- rbind(.ref_group, df)
  group <- factor(rep(c("ref", "x"), c(nrow(.ref_group), nrow(df))), levels = c("ref", "x"))

  df_cox <- data.frame(
    tte = data[[.var]],
    is_event = data[[is_event]],
    arm = group
  )
  if (is.null(strata)) {
    formula_cox <- survival::Surv(tte, is_event) ~ arm
  } else {
    formula_cox <- stats::as.formula(
      paste0(
        "survival::Surv(tte, is_event) ~ arm + strata(",
        paste(strata, collapse = ","),
        ")"
      )
    )
    df_cox <- cbind(df_cox, data[strata])
  }
  cox_fit <- survival::coxph(
    formula = formula_cox,
    data = df_cox,
    ties = ties
  )
  sum_cox <- summary(cox_fit, conf.int = conf_level, extend = TRUE)
  orginal_survdiff <- survival::survdiff(
    formula_cox,
    data = df_cox
  )
  log_rank_pvalue <- 1 - pchisq(orginal_survdiff$chisq, length(orginal_survdiff$n) - 1)

  pval <- switch(pval_method,
    "wald" = sum_cox$waldtest["pvalue"],
    "log-rank" = log_rank_pvalue, # pvalue from original log-rank test survival::survdiff()
    "likelihood" = sum_cox$logtest["pvalue"]
  )
  list(
    pvalue = formatters::with_label(unname(pval), paste0("p-value (", pval_method, ")")),
    hr = formatters::with_label(sum_cox$conf.int[1, 1], "Hazard Ratio"),
    hr_ci = formatters::with_label(unname(sum_cox$conf.int[1, 3:4]), f_conf_level(conf_level)),
    n_tot = formatters::with_label(sum_cox$n, "Total n"),
    n_tot_events = formatters::with_label(sum_cox$nevent, "Total events")
  )
}

#' @describeIn survival_coxph_pairwise Formatted analysis function which is used as `afun` in `coxph_pairwise()`.
#'
#' @return
#' * `a_coxph_pairwise()` returns the corresponding list with formatted [rtables::CellValue()].
#'
#' @keywords internal
a_coxph_pairwise <- make_afun(
  s_coxph_pairwise,
  .indent_mods = c(pvalue = 0L, hr = 0L, hr_ci = 1L, n_tot = 0L, n_tot_events = 0L),
  .formats = c(
    pvalue = "x.xxxx | (<0.0001)",
    hr = "xx.xx",
    hr_ci = "(xx.xx, xx.xx)",
    n_tot = "xx.xx",
    n_tot_events = "xx.xx"
  )
)

#' @describeIn survival_coxph_pairwise Layout-creating function which can take statistics function arguments
#'   and additional format arguments. This function is a wrapper for [rtables::analyze()].
#'
#' @return
#' * `coxph_pairwise()` returns a layout object suitable for passing to further layouting functions,
#'   or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
#'   the statistics from `s_coxph_pairwise()` to the table layout.
#'
#' @examples
#' library(dplyr)
#'
#' adtte_f <- tern_ex_adtte %>%
#'   filter(PARAMCD == "OS") %>%
#'   mutate(is_event = CNSR == 0)
#'
#' df <- adtte_f %>% filter(ARMCD == "ARM A")
#' df_ref_group <- adtte_f %>% filter(ARMCD == "ARM B")
#'
#' basic_table() %>%
#'   split_cols_by(var = "ARMCD", ref_group = "ARM A") %>%
#'   add_colcounts() %>%
#'   coxph_pairwise(
#'     vars = "AVAL",
#'     is_event = "is_event",
#'     var_labels = "Unstratified Analysis"
#'   ) %>%
#'   build_table(df = adtte_f)
#'
#' basic_table() %>%
#'   split_cols_by(var = "ARMCD", ref_group = "ARM A") %>%
#'   add_colcounts() %>%
#'   coxph_pairwise(
#'     vars = "AVAL",
#'     is_event = "is_event",
#'     var_labels = "Stratified Analysis",
#'     strata = "SEX",
#'     control = control_coxph(pval_method = "wald")
#'   ) %>%
#'   build_table(df = adtte_f)
#'
#' @export
#' @order 2
coxph_pairwise <- function(lyt,
                           vars,
                           strata = NULL,
                           control = control_coxph(),
                           na_str = default_na_str(),
                           nested = TRUE,
                           ...,
                           var_labels = "CoxPH",
                           show_labels = "visible",
                           table_names = vars,
                           .stats = c("pvalue", "hr", "hr_ci"),
                           .formats = NULL,
                           .labels = NULL,
                           .indent_mods = NULL) {
  extra_args <- list(strata = strata, control = control, ...)

  afun <- make_afun(
    a_coxph_pairwise,
    .stats = .stats,
    .formats = .formats,
    .labels = .labels,
    .indent_mods = .indent_mods
  )
  analyze(
    lyt,
    vars,
    var_labels = var_labels,
    show_labels = show_labels,
    table_names = table_names,
    afun = afun,
    na_str = na_str,
    nested = nested,
    extra_args = extra_args
  )
}

#' Count number of patients with missed doses by thresholds
#'
#' @description `r lifecycle::badge("stable")`
#'
#' The analyze function creates a layout element to calculate cumulative counts of patients with number of missed
#' doses at least equal to user-specified threshold values.
#'
#' This function analyzes numeric variable `vars`, a variable with numbers of missed doses,
#' against the threshold values supplied to the `thresholds` argument as a numeric vector. This function
#' assumes that every row of the given data frame corresponds to a unique patient.
#'
#' @inheritParams s_count_cumulative
#' @inheritParams argument_convention
#' @param thresholds (`numeric`)\cr minimum number of missed doses the patients had.
#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("count_missed_doses")`
#'   to see available statistics for this function.
#'
#' @seealso
#' * Relevant description function [d_count_missed_doses()] which generates labels for [count_missed_doses()].
#' * Similar analyze function [count_cumulative()] which more generally counts cumulative values and has more
#'   options for threshold handling, but uses different labels.
#'
#' @name count_missed_doses
#' @order 1
NULL

#' @describeIn count_missed_doses Statistics function to count non-missing values.
#'
#' @return
#' * `s_count_nonmissing()` returns the statistic `n` which is the count of non-missing values in `x`.
#'
#' @keywords internal
s_count_nonmissing <- function(x) {
  list(n = n_available(x))
}

#' Description function that calculates labels for `s_count_missed_doses()`
#'
#' @description `r lifecycle::badge("stable")`
#'
#' @inheritParams s_count_missed_doses
#'
#' @return [d_count_missed_doses()] returns a named `character` vector with the labels.
#'
#' @seealso [s_count_missed_doses()]
#'
#' @export
d_count_missed_doses <- function(thresholds) {
  paste0("At least ", thresholds, " missed dose", ifelse(thresholds > 1, "s", ""))
}

#' @describeIn count_missed_doses Statistics function to count patients with missed doses.
#'
#' @return
#' * `s_count_missed_doses()` returns the statistics `n` and `count_fraction` with one element for each threshold.
#'
#' @keywords internal
s_count_missed_doses <- function(x,
                                 thresholds,
                                 .N_col) { # nolint
  stat <- s_count_cumulative(
    x = x,
    thresholds = thresholds,
    lower_tail = FALSE,
    include_eq = TRUE,
    .N_col = .N_col
  )
  labels <- d_count_missed_doses(thresholds)
  for (i in seq_along(stat$count_fraction)) {
    stat$count_fraction[[i]] <- formatters::with_label(stat$count_fraction[[i]], label = labels[i])
  }
  n_stat <- s_count_nonmissing(x)
  c(n_stat, stat)
}

#' @describeIn count_missed_doses Formatted analysis function which is used as `afun`
#'   in `count_missed_doses()`.
#'
#' @return
#' * `a_count_missed_doses()` returns the corresponding list with formatted [rtables::CellValue()].
#'
#' @keywords internal
a_count_missed_doses <- make_afun(
  s_count_missed_doses,
  .formats = c(n = "xx", count_fraction = format_count_fraction)
)

#' @describeIn count_missed_doses Layout-creating function which can take statistics function arguments
#'   and additional format arguments. This function is a wrapper for [rtables::analyze()].
#'
#' @return
#' * `count_missed_doses()` returns a layout object suitable for passing to further layouting functions,
#'   or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
#'   the statistics from `s_count_missed_doses()` to the table layout.
#'
#' @examples
#' library(dplyr)
#'
#' anl <- tern_ex_adsl %>%
#'   distinct(STUDYID, USUBJID, ARM) %>%
#'   mutate(
#'     PARAMCD = "TNDOSMIS",
#'     PARAM = "Total number of missed doses during study",
#'     AVAL = sample(0:20, size = nrow(tern_ex_adsl), replace = TRUE),
#'     AVALC = ""
#'   )
#'
#' basic_table() %>%
#'   split_cols_by("ARM") %>%
#'   add_colcounts() %>%
#'   count_missed_doses("AVAL", thresholds = c(1, 5, 10, 15), var_labels = "Missed Doses") %>%
#'   build_table(anl, alt_counts_df = tern_ex_adsl)
#'
#' @export
#' @order 2
count_missed_doses <- function(lyt,
                               vars,
                               thresholds,
                               var_labels = vars,
                               show_labels = "visible",
                               na_str = default_na_str(),
                               nested = TRUE,
                               ...,
                               table_names = vars,
                               .stats = NULL,
                               .formats = NULL,
                               .labels = NULL,
                               .indent_mods = NULL) {
  extra_args <- list(thresholds = thresholds, ...)

  afun <- make_afun(
    a_count_missed_doses,
    .stats = .stats,
    .formats = .formats,
    .labels = .labels,
    .indent_mods = .indent_mods,
    .ungroup_stats = "count_fraction"
  )
  analyze(
    lyt = lyt,
    vars = vars,
    afun = afun,
    var_labels = var_labels,
    table_names = table_names,
    show_labels = show_labels,
    na_str = na_str,
    nested = nested,
    extra_args = extra_args
  )
}

#' Count specific values
#'
#' @description `r lifecycle::badge("stable")`
#'
#' The analyze function [count_values()] creates a layout element to calculate counts of specific values within a
#' variable of interest.
#'
#' This function analyzes one or more variables of interest supplied as a vector to `vars`. Values to
#' count for variable(s) in `vars` can be given as a vector via the `values` argument. One row of
#' counts will be generated for each variable.
#'
#' @inheritParams argument_convention
#' @param values (`character`)\cr specific values that should be counted.
#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("count_values")`
#'   to see available statistics for this function.
#'
#' @note
#' * For `factor` variables, `s_count_values` checks whether `values` are all included in the levels of `x`
#'   and fails otherwise.
#' * For `count_values()`, variable labels are shown when there is more than one element in `vars`,
#'   otherwise they are hidden.
#'
#' @name count_values
#' @order 1
NULL

#' @describeIn count_values S3 generic function to count values.
#'
#' @inheritParams s_summary.logical
#'
#' @return
#' * `s_count_values()` returns output of [s_summary()] for specified values of a non-numeric variable.
#'
#' @export
s_count_values <- function(x,
                           values,
                           na.rm = TRUE, # nolint
                           .N_col, # nolint
                           .N_row, # nolint
                           denom = c("n", "N_row", "N_col")) {
  UseMethod("s_count_values", x)
}

#' @describeIn count_values Method for `character` class.
#'
#' @method s_count_values character
#'
#' @examples
#' # `s_count_values.character`
#' s_count_values(x = c("a", "b", "a"), values = "a")
#' s_count_values(x = c("a", "b", "a", NA, NA), values = "b", na.rm = FALSE)
#'
#' @export
s_count_values.character <- function(x,
                                     values = "Y",
                                     na.rm = TRUE, # nolint
                                     ...) {
  checkmate::assert_character(values)

  if (na.rm) {
    x <- x[!is.na(x)]
  }

  is_in_values <- x %in% values

  s_summary(is_in_values, ...)
}

#' @describeIn count_values Method for `factor` class. This makes an automatic
#'   conversion to `character` and then forwards to the method for characters.
#'
#' @method s_count_values factor
#'
#' @examples
#' # `s_count_values.factor`
#' s_count_values(x = factor(c("a", "b", "a")), values = "a")
#'
#' @export
s_count_values.factor <- function(x,
                                  values = "Y",
                                  ...) {
  s_count_values(as.character(x), values = as.character(values), ...)
}

#' @describeIn count_values Method for `logical` class.
#'
#' @method s_count_values logical
#'
#' @examples
#' # `s_count_values.logical`
#' s_count_values(x = c(TRUE, FALSE, TRUE))
#'
#' @export
s_count_values.logical <- function(x, values = TRUE, ...) {
  checkmate::assert_logical(values)
  s_count_values(as.character(x), values = as.character(values), ...)
}

#' @describeIn count_values Formatted analysis function which is used as `afun`
#'   in `count_values()`.
#'
#' @return
#' * `a_count_values()` returns the corresponding list with formatted [rtables::CellValue()].
#'
#' @examples
#' # `a_count_values`
#' a_count_values(x = factor(c("a", "b", "a")), values = "a", .N_col = 10, .N_row = 10)
#'
#' @export
a_count_values <- make_afun(
  s_count_values,
  .formats = c(count_fraction = "xx (xx.xx%)", count = "xx")
)

#' @describeIn count_values Layout-creating function which can take statistics function arguments
#'   and additional format arguments. This function is a wrapper for [rtables::analyze()].
#'
#' @return
#' * `count_values()` returns a layout object suitable for passing to further layouting functions,
#'   or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
#'   the statistics from `s_count_values()` to the table layout.
#'
#' @examples
#' # `count_values`
#' basic_table() %>%
#'   count_values("Species", values = "setosa") %>%
#'   build_table(iris)
#'
#' @export
#' @order 2
count_values <- function(lyt,
                         vars,
                         values,
                         na_str = default_na_str(),
                         nested = TRUE,
                         ...,
                         table_names = vars,
                         .stats = "count_fraction",
                         .formats = NULL,
                         .labels = c(count_fraction = paste(values, collapse = ", ")),
                         .indent_mods = NULL) {
  extra_args <- list(values = values, ...)

  afun <- make_afun(
    a_count_values,
    .stats = .stats,
    .formats = .formats,
    .labels = .labels,
    .indent_mods = .indent_mods
  )
  analyze(
    lyt,
    vars,
    afun = afun,
    na_str = na_str,
    nested = nested,
    extra_args = extra_args,
    show_labels = ifelse(length(vars) > 1, "visible", "hidden"),
    table_names = table_names
  )
}

#' Control function for Cox-PH model
#'
#' @description `r lifecycle::badge("stable")`
#'
#' This is an auxiliary function for controlling arguments for Cox-PH model, typically used internally to specify
#' details of Cox-PH model for [s_coxph_pairwise()]. `conf_level` refers to Hazard Ratio estimation.
#'
#' @inheritParams argument_convention
#' @param pval_method (`string`)\cr p-value method for testing hazard ratio = 1.
#'   Default method is `"log-rank"`, can also be set to `"wald"` or `"likelihood"`.
#' @param ties (`string`)\cr string specifying the method for tie handling. Default is `"efron"`,
#'   can also be set to `"breslow"` or `"exact"`. See more in [survival::coxph()].
#'
#' @return A list of components with the same names as the arguments.
#'
#' @export
control_coxph <- function(pval_method = c("log-rank", "wald", "likelihood"),
                          ties = c("efron", "breslow", "exact"),
                          conf_level = 0.95) {
  pval_method <- match.arg(pval_method)
  ties <- match.arg(ties)
  assert_proportion_value(conf_level)

  list(pval_method = pval_method, ties = ties, conf_level = conf_level)
}

#' Control function for `survfit` models for survival time
#'
#' @description `r lifecycle::badge("stable")`
#'
#' This is an auxiliary function for controlling arguments for `survfit` model, typically used internally to specify
#' details of `survfit` model for [s_surv_time()]. `conf_level` refers to survival time estimation.
#'
#' @inheritParams argument_convention
#' @param conf_type (`string`)\cr confidence interval type. Options are "plain" (default), "log", "log-log",
#'   see more in [survival::survfit()]. Note option "none" is no longer supported.
#' @param quantiles (`numeric(2)`)\cr vector of length two specifying the quantiles of survival time.
#'
#' @return A list of components with the same names as the arguments.
#'
#' @export
control_surv_time <- function(conf_level = 0.95,
                              conf_type = c("plain", "log", "log-log"),
                              quantiles = c(0.25, 0.75)) {
  conf_type <- match.arg(conf_type)
  checkmate::assert_numeric(quantiles, lower = 0, upper = 1, len = 2, unique = TRUE, sorted = TRUE)
  nullo <- lapply(quantiles, assert_proportion_value)
  assert_proportion_value(conf_level)
  list(conf_level = conf_level, conf_type = conf_type, quantiles = quantiles)
}

#' Control function for `survfit` models for patients' survival rate at time points
#'
#' @description `r lifecycle::badge("stable")`
#'
#' This is an auxiliary function for controlling arguments for `survfit` model, typically used internally to specify
#' details of `survfit` model for [s_surv_timepoint()]. `conf_level` refers to patient risk estimation at a time point.
#'
#' @inheritParams argument_convention
#' @inheritParams control_surv_time
#'
#' @return A list of components with the same names as the arguments.
#'
#' @export
control_surv_timepoint <- function(conf_level = 0.95,
                                   conf_type = c("plain", "log", "log-log")) {
  conf_type <- match.arg(conf_type)
  assert_proportion_value(conf_level)
  list(
    conf_level = conf_level,
    conf_type = conf_type
  )
}

#' Helper function for tabulation of a single biomarker result
#'
#' @description `r lifecycle::badge("stable")`
#'
#' Please see [h_tab_surv_one_biomarker()] and [h_tab_rsp_one_biomarker()], which use this function for examples.
#' This function is a wrapper for [rtables::summarize_row_groups()].
#'
#' @inheritParams argument_convention
#' @param df (`data.frame`)\cr results for a single biomarker.
#' @param afuns (named `list` of `function`)\cr analysis functions.
#' @param colvars (named `list`)\cr named list with elements `vars` (variables to tabulate) and `labels` (their labels).
#'
#' @return An `rtables` table object with statistics in columns.
#'
#' @export
h_tab_one_biomarker <- function(df,
                                afuns,
                                colvars,
                                na_str = default_na_str(),
                                .indent_mods = 0L,
                                ...) {
  extra_args <- list(...)

  # Create "ci" column from "lcl" and "ucl"
  df$ci <- combine_vectors(df$lcl, df$ucl)

  lyt <- basic_table()

  # Row split by row type - only keep the content rows here.
  lyt <- split_rows_by(
    lyt = lyt,
    var = "row_type",
    split_fun = keep_split_levels("content"),
    nested = FALSE
  )

  # Summarize rows with all patients.
  lyt <- summarize_row_groups(
    lyt = lyt,
    var = "var_label",
    cfun = afuns,
    na_str = na_str,
    indent_mod = .indent_mods,
    extra_args = extra_args
  )

  # Split cols by the multiple variables to populate into columns.
  lyt <- split_cols_by_multivar(
    lyt = lyt,
    vars = colvars$vars,
    varlabels = colvars$labels
  )

  # If there is any subgroup variables, we extend the layout accordingly.
  if ("analysis" %in% df$row_type) {
    # Now only continue with the subgroup rows.
    lyt <- split_rows_by(
      lyt = lyt,
      var = "row_type",
      split_fun = keep_split_levels("analysis"),
      nested = FALSE,
      child_labels = "hidden"
    )

    # Split by the subgroup variable.
    lyt <- split_rows_by(
      lyt = lyt,
      var = "var",
      labels_var = "var_label",
      nested = TRUE,
      child_labels = "visible",
      indent_mod = .indent_mods * 2
    )

    # Then analyze colvars for each subgroup.
    lyt <- summarize_row_groups(
      lyt = lyt,
      cfun = afuns,
      var = "subgroup",
      na_str = na_str,
      extra_args = extra_args
    )
  }
  build_table(lyt, df = df)
}

#' Count the number of patients with a particular event
#'
#' @description `r lifecycle::badge("stable")`
#'
#' The analyze function [count_patients_with_event()] creates a layout element to calculate patient counts for a
#' user-specified set of events.
#'
#' This function analyzes primary analysis variable `vars` which indicates unique subject identifiers. Events
#' are defined by the user as a named vector via the `filters` argument, where each name corresponds to a
#' variable and each value is the value(s) that that variable takes for the event.
#'
#' If there are multiple records with the same event recorded for a patient, only one occurrence is counted.
#'
#' @inheritParams argument_convention
#' @param filters (`character`)\cr a character vector specifying the column names and flag variables
#'   to be used for counting the number of unique identifiers satisfying such conditions.
#'   Multiple column names and flags are accepted in this format
#'   `c("column_name1" = "flag1", "column_name2" = "flag2")`.
#'   Note that only equality is being accepted as condition.
#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("count_patients_with_event")`
#'   to see available statistics for this function.
#'
#' @seealso [count_patients_with_flags]
#'
#' @name count_patients_with_event
#' @order 1
NULL

#' @describeIn count_patients_with_event Statistics function which counts the number of patients for which
#'   the defined event has occurred.
#'
#' @inheritParams analyze_variables
#' @param .var (`string`)\cr name of the column that contains the unique identifier.
#'
#' @return
#' * `s_count_patients_with_event()` returns the count and fraction of unique identifiers with the defined event.
#'
#' @examples
#' # `s_count_patients_with_event()`
#'
#' s_count_patients_with_event(
#'   tern_ex_adae,
#'   .var = "SUBJID",
#'   filters = c("TRTEMFL" = "Y")
#' )
#'
#' s_count_patients_with_event(
#'   tern_ex_adae,
#'   .var = "SUBJID",
#'   filters = c("TRTEMFL" = "Y", "AEOUT" = "FATAL")
#' )
#'
#' s_count_patients_with_event(
#'   tern_ex_adae,
#'   .var = "SUBJID",
#'   filters = c("TRTEMFL" = "Y", "AEOUT" = "FATAL"),
#'   denom = "N_col",
#'   .N_col = 456
#' )
#'
#' @export
s_count_patients_with_event <- function(df,
                                        .var,
                                        filters,
                                        .N_col, # nolint
                                        .N_row, # nolint
                                        denom = c("n", "N_row", "N_col")) {
  col_names <- names(filters)
  filter_values <- filters

  checkmate::assert_subset(col_names, colnames(df))

  temp <- Map(
    function(x, y) which(df[[x]] == y),
    col_names,
    filter_values
  )
  position_satisfy_filters <- Reduce(intersect, temp)
  id_satisfy_filters <- as.character(unique(df[position_satisfy_filters, ][[.var]]))
  result <- s_count_values(
    as.character(unique(df[[.var]])),
    id_satisfy_filters,
    denom = denom,
    .N_col = .N_col,
    .N_row = .N_row
  )
  result
}

#' @describeIn count_patients_with_event Formatted analysis function which is used as `afun`
#'   in `count_patients_with_event()`.
#'
#' @return
#' * `a_count_patients_with_event()` returns the corresponding list with formatted [rtables::CellValue()].
#'
#' @examples
#' # `a_count_patients_with_event()`
#'
#' a_count_patients_with_event(
#'   tern_ex_adae,
#'   .var = "SUBJID",
#'   filters = c("TRTEMFL" = "Y"),
#'   .N_col = 100,
#'   .N_row = 100
#' )
#'
#' @export
a_count_patients_with_event <- make_afun(
  s_count_patients_with_event,
  .formats = c(count_fraction = format_count_fraction_fixed_dp)
)

#' @describeIn count_patients_with_event Layout-creating function which can take statistics function
#'   arguments and additional format arguments. This function is a wrapper for [rtables::analyze()].
#'
#' @return
#' * `count_patients_with_event()` returns a layout object suitable for passing to further layouting functions,
#'   or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
#'   the statistics from `s_count_patients_with_event()` to the table layout.
#'
#' @examples
#' # `count_patients_with_event()`
#'
#' lyt <- basic_table() %>%
#'   split_cols_by("ARM") %>%
#'   add_colcounts() %>%
#'   count_values(
#'     "STUDYID",
#'     values = "AB12345",
#'     .stats = "count",
#'     .labels = c(count = "Total AEs")
#'   ) %>%
#'   count_patients_with_event(
#'     "SUBJID",
#'     filters = c("TRTEMFL" = "Y"),
#'     .labels = c(count_fraction = "Total number of patients with at least one adverse event"),
#'     table_names = "tbl_all"
#'   ) %>%
#'   count_patients_with_event(
#'     "SUBJID",
#'     filters = c("TRTEMFL" = "Y", "AEOUT" = "FATAL"),
#'     .labels = c(count_fraction = "Total number of patients with fatal AEs"),
#'     table_names = "tbl_fatal"
#'   ) %>%
#'   count_patients_with_event(
#'     "SUBJID",
#'     filters = c("TRTEMFL" = "Y", "AEOUT" = "FATAL", "AEREL" = "Y"),
#'     .labels = c(count_fraction = "Total number of patients with related fatal AEs"),
#'     .indent_mods = c(count_fraction = 2L),
#'     table_names = "tbl_rel_fatal"
#'   )
#'
#' build_table(lyt, tern_ex_adae, alt_counts_df = tern_ex_adsl)
#'
#' @export
#' @order 2
count_patients_with_event <- function(lyt,
                                      vars,
                                      filters,
                                      riskdiff = FALSE,
                                      na_str = default_na_str(),
                                      nested = TRUE,
                                      ...,
                                      table_names = vars,
                                      .stats = "count_fraction",
                                      .formats = NULL,
                                      .labels = NULL,
                                      .indent_mods = NULL) {
  checkmate::assert_flag(riskdiff)

  s_args <- list(filters = filters, ...)

  afun <- make_afun(
    a_count_patients_with_event,
    .stats = .stats,
    .formats = .formats,
    .labels = .labels,
    .indent_mods = .indent_mods
  )

  extra_args <- if (isFALSE(riskdiff)) {
    s_args
  } else {
    list(
      afun = list("s_count_patients_with_event" = afun),
      .stats = .stats,
      .indent_mods = .indent_mods,
      s_args = s_args
    )
  }

  analyze(
    lyt,
    vars,
    afun = ifelse(isFALSE(riskdiff), afun, afun_riskdiff),
    na_str = na_str,
    nested = nested,
    extra_args = extra_args,
    show_labels = ifelse(length(vars) > 1, "visible", "hidden"),
    table_names = table_names
  )
}

#' Control function for logistic regression model fitting
#'
#' @description `r lifecycle::badge("stable")`
#'
#' This is an auxiliary function for controlling arguments for logistic regression models.
#' `conf_level` refers to the confidence level used for the Odds Ratio CIs.
#'
#' @inheritParams argument_convention
#' @param response_definition (`string`)\cr the definition of what an event is in terms of `response`.
#'   This will be used when fitting the logistic regression model on the left hand side of the formula.
#'   Note that the evaluated expression should result in either a logical vector or a factor with 2
#'   levels. By default this is just `"response"` such that the original response variable is used
#'   and not modified further.
#'
#' @return A list of components with the same names as the arguments.
#'
#' @examples
#' # Standard options.
#' control_logistic()
#'
#' # Modify confidence level.
#' control_logistic(conf_level = 0.9)
#'
#' # Use a different response definition.
#' control_logistic(response_definition = "I(response %in% c('CR', 'PR'))")
#'
#' @export
control_logistic <- function(response_definition = "response",
                             conf_level = 0.95) {
  checkmate::assert_true(grepl("response", response_definition))
  checkmate::assert_string(response_definition)
  assert_proportion_value(conf_level)
  list(
    response_definition = response_definition,
    conf_level = conf_level
  )
}

#' Control function for incidence rate
#'
#' @description `r lifecycle::badge("stable")`
#'
#' This is an auxiliary function for controlling arguments for the incidence rate, used
#' internally to specify details in `s_incidence_rate()`.
#'
#' @inheritParams argument_convention
#' @param conf_type (`string`)\cr `normal` (default), `normal_log`, `exact`, or `byar`
#'   for confidence interval type.
#' @param input_time_unit (`string`)\cr `day`, `week`, `month`, or `year` (default)
#'   indicating time unit for data input.
#' @param num_pt_year (`numeric(1)`)\cr number of patient-years to use when calculating adverse event rates.
#'
#' @return A list of components with the same names as the arguments.
#'
#' @seealso [incidence_rate]
#'
#' @examples
#' control_incidence_rate(0.9, "exact", "month", 100)
#'
#' @export
control_incidence_rate <- function(conf_level = 0.95,
                                   conf_type = c("normal", "normal_log", "exact", "byar"),
                                   input_time_unit = c("year", "day", "week", "month"),
                                   num_pt_year = 100) {
  conf_type <- match.arg(conf_type)
  input_time_unit <- match.arg(input_time_unit)
  checkmate::assert_number(num_pt_year)
  assert_proportion_value(conf_level)

  list(
    conf_level = conf_level,
    conf_type = conf_type,
    input_time_unit = input_time_unit,
    num_pt_year = num_pt_year
  )
}

#' Generate PK reference dataset
#'
#' @description `r lifecycle::badge("stable")`
#'
#' @return A `data.frame` of PK parameters.
#'
#' @examples
#' pk_reference_dataset <- d_pkparam()
#'
#' @export
d_pkparam <- function() {
  pk_dataset <- as.data.frame(matrix(
    c(
      "TMAX", "Time of CMAX", "Tmax", "Plasma/Blood/Serum", "1",
      "CMAX", "Max Conc", "Cmax", "Plasma/Blood/Serum", "2",
      "CMAXD", "Max Conc Norm by Dose", "Cmax/D", "Plasma/Blood/Serum", "3",
      "AUCIFO", "AUC Infinity Obs", "AUCinf obs", "Plasma/Blood/Serum", "4",
      "AUCIFP", "AUC Infinity Pred", "AUCinf pred", "Plasma/Blood/Serum", "5",
      "AUCIFOD", "AUC Infinity Obs Norm by Dose", "AUCinf/D obs", "Plasma/Blood/Serum", "6",
      "AUCIFD", "AUC Infinity Pred Norm by Dose", "AUCinf/D pred", "Plasma/Blood/Serum", "7",
      "AUCPEO", "AUC %Extrapolation Obs", "AUCinf extrap obs", "Plasma/Blood/Serum", "8",
      "AUCPEP", "AUC %Extrapolation Pred", "AUCinf extrap pred", "Plasma/Blood/Serum", "9",
      "AUCINT", "AUC from T1 to T2", "AUCupper-lower ", "Plasma/Blood/Serum", "10",
      "AUCTAU", "AUC Over Dosing Interval", "AUCtau", "Plasma/Blood/Serum", "11",
      "AUCLST", "AUC to Last Nonzero Conc", "AUClast", "Plasma/Blood/Serum", "12",
      "AUCALL", "AUC All", "AUCall", "Plasma/Blood/Serum", "13",
      "AUMCIFO", "AUMC Infinity Obs", "AUMCinf obs", "Plasma/Blood/Serum", "14",
      "AUMCIFP", "AUMC Infinity Pred", "AUMCinf pred", "Plasma/Blood/Serum", "15",
      "AUMCPEO", "AUMC % Extrapolation Obs", "AUMC extrap obs", "Plasma/Blood/Serum", "16",
      "AUMCPEP", "AUMC % Extrapolation Pred", "AUMC extrap pred", "Plasma/Blood/Serum", "17",
      "AUMCTAU", "AUMC Over Dosing Interval", "AUMCtau", "Plasma/Blood/Serum", "18",
      "AUMCLST", "AUMC to Last Nonzero Conc", "AUMClast", "Plasma/Blood/Serum", "19",
      "AURCIFO", "AURC Infinity Obs", "AURCinf obs", "Plasma/Blood/Serum", "20",
      "AURCIFP", "AURC Infinity Pred", "AURCinf pred", "Plasma/Blood/Serum", "21",
      "AURCPEO", "AURC % Extrapolation Obs", "AURC extrap obs", "Plasma/Blood/Serum", "22",
      "AURCPEP", "AURC % Extrapolation Pred", "AURC extrap pred", "Plasma/Blood/Serum", "23",
      "AURCLST", "AURC Dosing to Last Conc", "AURClast", "Plasma/Blood/Serum", "24",
      "AURCALL", "AURC All", "AURCall", "Plasma/Blood/Serum", "25",
      "TLST", "Time of Last Nonzero Conc", "Tlast", "Plasma/Blood/Serum", "26",
      "CO", "Initial Conc", "CO", "Plasma/Blood/Serum", "27",
      "C0", "Initial Conc", "C0", "Plasma/Blood/Serum", "28",
      "CAVG", "Average Conc", "Cavg", "Plasma/Blood/Serum", "29",
      "CLST", "Last Nonzero Conc", "Clast", "Plasma/Blood/Serum", "30",
      "CMIN", "Min Conc", "Cmin", "Plasma/Blood/Serum", "31",
      "LAMZHL", "Half-Life Lambda z", "t1/2", "Plasma/Blood/Serum", "32",
      "CLFO", "Total CL Obs by F", "CL/F obs", "Plasma/Blood/Serum", "33",
      "CLFP", "Total CL Pred by F", "CL/F pred", "Plasma/Blood/Serum", "34",
      "CLO", "Total CL Obs", "CL obs", "Plasma/Blood/Serum", "35",
      "CLP", "Total CL Pred", "CL pred", "Plasma/Blood/Serum", "36",
      "CLSS", "Total CL Steady State Pred", "CLss", "Plasma/Blood/Serum", "37",
      "CLSSF", "Total CL Steady State Pred by F", "CLss/F", "Plasma/Blood/Serum", "38",
      "VZFO", "Vz Obs by F", "Vz/F obs", "Plasma/Blood/Serum", "39",
      "VZFP", "Vz Pred by F", "Vz/F pred", "Plasma/Blood/Serum", "40",
      "VZO", "Vz Obs", "Vz obs", "Plasma/Blood/Serum", "41",
      "VZP", "Vz Pred", "Vz pred", "Plasma/Blood/Serum", "42",
      "VSSO", "Vol Dist Steady State Obs", "Vss obs", "Plasma/Blood/Serum", "43",
      "VSSP", "Vol Dist Steady State Pred", "Vss pred", "Plasma/Blood/Serum", "44",
      "LAMZ", "Lambda z", "Lambda z", "Plasma/Blood/Serum", "45",
      "LAMZLL", "Lambda z Lower Limit", "Lambda z lower", "Plasma/Blood/Serum", "46",
      "LAMZUL", "Lambda z Upper Limit", "Lambda z upper", "Plasma/Blood/Serum", "47",
      "LAMZNPT", "Number of Points for Lambda z", "No points Lambda z", "Plasma/Blood/Serum", "48",
      "MRTIFO", "MRT Infinity Obs", "MRTinf obs", "Plasma/Blood/Serum", "49",
      "MRTIFP", "MRT Infinity Pred", "MRTinf pred", "Plasma/Blood/Serum", "50",
      "MRTLST", "MRT to Last Nonzero Conc", "MRTlast", "Plasma/Blood/Serum", "51",
      "R2", "R Squared", "Rsq", "Plasma/Blood/Serum", "52",
      "R2ADJ", "R Squared Adjusted", "Rsq adjusted", "Plasma/Blood/Serum", "53",
      "TLAG", "Time Until First Nonzero Conc", "TIag", "Plasma/Blood/Serum", "54",
      "TMIN", "Time of CMIN Observation", "Tmin", "Plasma/Blood/Serum", "55",
      "ACCI", "Accumulation Index", "Accumulation Index", "Plasma/Blood/Serum/Urine", "56",
      "FLUCP", "Fluctuation%", "Fluctuation", "Plasma/Blood/Serum", "57",
      "CORRXY", "Correlation Between TimeX and Log ConcY", "Corr xy", "Plasma/Blood/Serum", "58",
      "RCAMINT", "Amt Rec from T1 to T2", "Ae", "Urine", "59",
      "RCPCINT", "Pct Rec from T1 to T2", "Fe", "Urine", "60",
      "VOLPK", "Sum of Urine Vol", "Urine volume", "Urine", "61",
      "RENALCL", "Renal CL", "CLR", "Plasma/Blood/Serum/Urine", "62",
      "ERTMAX", "Time of Max Excretion Rate", "Tmax Rate", "Urine", "63",
      "RMAX", "Time of Maximum Response", "Rmax", "Matrix of PD", "64",
      "RMIN", "Time of Minimum Response", "Rmin", "Matrix of PD", "65",
      "ERMAX", "Max Excretion Rate", "Max excretion rate", "Urine", "66",
      "MIDPTLST", "Midpoint of Collection Interval", "Midpoint last", "Urine", "67",
      "ERLST", "Last Meas Excretion Rate", "Rate last", "Urine", "68",
      "TON", "Time to Onset", "Tonset", "Matrix of PD", "69",
      "TOFF", "Time to Offset", "Toffset", "Matrix of PD", "70",
      "TBBLP", "Time Below Baseline %", "Time %Below Baseline", "Matrix of PD", "71",
      "TBTP", "Time Below Threshold %", "Time %Below Threshold", "Matrix of PD", "72",
      "TABL", "Time Above Baseline", "Time Above Baseline", "Matrix of PD", "73",
      "TAT", "Time Above Threshold", "Time Above Threshold", "Matrix of PD", "74",
      "TBT", "Time Below Threshold", "Time Below Threshold", "Matrix of PD", "75",
      "TBLT", "Time Between Baseline and Threshold", "Time Between Baseline Threshold", "Matrix of PD", "76",
      "BLRSP", "Baseline Response", "Baseline", "Matrix of PD", "77",
      "TSHDRSP", "Response Threshold", "Threshold", "Matrix of PD", "78",
      "AUCABL", "AUC Above Baseline", "AUC above baseline", "Matrix of PD", "79",
      "AUCAT", "AUC Above Threshold", "AUC above threshold", "Matrix of PD", "80",
      "AUCBBL", "AUC Below Baseline", "AUC below baseline", "Matrix of PD", "81",
      "AUCBT", "AUC Below Threshold", "AUC below threshold", "Matrix of PD", "82",
      "AUCBLDIF", "Diff AUC Above Base and AUC Below Base", "AUC diff baseline", "Matrix of PD", "83",
      "AUCTDIF", "Diff AUC Above Thr and AUC Below Thr", "AUCnet threshold", "Matrix of PD", "84",
      "TDIFF", "Diff Time to Offset and Time to Onset", "Diff toffset-tonset", "Matrix of PD", "85",
      "AUCPBEO", "AUC %Back Extrapolation Obs", "AUC%Back extrap obs", "Plasma/Blood/Serum", "86",
      "AUCPBEP", "AUC %Back Extrapolation Pred", "AUC%Back extrap pred", "Plasma/Blood/Serum", "87",
      "TSLP1L", "Lower Time Limit Slope 1st", "Slope1 lower", "Matrix of PD", "88",
      "TSLP1U", "Upper Time Limit Slope 1st Segment", "Slope1 upper", "Matrix of PD", "89",
      "TSLP2L", "Lower Time Limit Slope 2nd Segment", "Slope2 lower", "Matrix of PD", "90",
      "TSLP2U", "Upper Time Limit Slope 2nd Segment", "Slope2 upper", "Matrix of PD", "91",
      "SLP1", "Slope, 1st Segment", "Slope1", "Matrix of PD", "92",
      "SLP2", "Slope, 2nd Segment", "Slope2", "Matrix of PD", "93",
      "SLP1PT", "Number of Points for Slope 1st Segment", "No points slope1", "Matrix of PD", "94",
      "SLP2PT", "Number of Points for Slope 2nd Segment", "No points slope2", "Matrix of PD", "95",
      "R2ADJS1", "R-Squared Adjusted Slope, 1st Segment", "Rsq adjusted slope1", "Matrix of PD", "96",
      "R2ADJS2", "R-Squared Adjusted Slope, 2nd Segment", "Rsq adjusted slope2", "Matrix of PD", "97",
      "R2SLP1", "R Squared, Slope, 1st Segment", "Rsq slope1", "Matrix of PD", "98",
      "R2SLP2", "R Squared, Slope, 2nd Segment", "Rsq slope2", "Matrix of PD", "99",
      "CORRXYS1", "Corr Btw TimeX and Log ConcY, Slope 1st", "Corr xy slope1", "Plasma/Blood/Serum", "100",
      "CORRXYS2", "Corr Btw TimeX and Log ConcY, Slope 1st Slope 2nd", "Corr xy slope2", "Plasma/Blood/Serum", "101",
      "AILAMZ", "Accumulation Index using Lambda z", "AILAMZ", "Plasma/Blood/Serum", "102",
      "ARAUC", "Accumulation Ratio AUCTAU", "ARAUC", "Plasma/Blood/Serum", "103",
      "ARAUCD", "Accum Ratio AUCTAU norm by dose", "ARAUCD", "Plasma/Blood/Serum", "104",
      "ARAUCIFO", "Accum Ratio AUC Infinity Obs", "ARAUCIFO", "Plasma/Blood/Serum", "105",
      "ARAUCIFP", "Accum Ratio AUC Infinity Pred", "ARAUCIFP", "Plasma/Blood/Serum", "106",
      "ARAUCIND", "Accum Ratio AUC T1 to T2 norm by dose", "ARAUCIND_T1_T2_UNIT", "Plasma/Blood/Serum", "107",
      "ARAUCINT", "Accumulation Ratio AUC from T1 to T2", "ARAUCINT_T1_T2_UNIT", "Plasma/Blood/Serum", "108",
      "ARAUCIOD", "Accum Ratio AUCIFO Norm by Dose", "ARAUCIOD", "Plasma/Blood/Serum", "109",
      "ARAUCIPD", "Accum Ratio AUCIFP Norm by Dose", "ARAUCIPD", "Plasma/Blood/Serum", "110",
      "ARAUCLST", "Accum Ratio AUC to Last Nonzero Conc", "ARAUCLST", "Plasma/Blood/Serum", "111",
      "ARCMAX", "Accumulation Ratio Cmax", "ARCMAX", "Plasma/Blood/Serum", "112",
      "ARCMAXD", "Accum Ratio Cmax norm by dose", "ARCMAXD", "Plasma/Blood/Serum", "113",
      "ARCMIN", "Accumulation Ratio Cmin", "ARCMIN", "Plasma/Blood/Serum", "114",
      "ARCMIND", "Accum Ratio Cmin norm by dose", "ARCMIND", "Plasma/Blood/Serum", "115",
      "ARCTROUD", "Accum Ratio Ctrough norm by dose", "ARCTROUD", "Plasma/Blood/Serum", "116",
      "ARCTROUG", "Accumulation Ratio Ctrough", "ARCTROUG", "Plasma/Blood/Serum", "117",
      "AUCALLB", "AUC All Norm by BMI", "AUCall_B", "Plasma/Blood/Serum", "118",
      "AUCALLD", "AUC All Norm by Dose", "AUCall_D", "Plasma/Blood/Serum", "119",
      "AUCALLS", "AUC All Norm by SA", "AUCall_S", "Plasma/Blood/Serum", "120",
      "AUCALLW", "AUC All Norm by WT", "AUCall_W", "Plasma/Blood/Serum", "121",
      "AUCIFOB", "AUC Infinity Obs Norm by BMI", "AUCINF_obs_B", "Plasma/Blood/Serum", "122",
      "AUCIFOLN", "AUC Infinity Obs LN Transformed", "AUCIFOLN", "Plasma/Blood/Serum", "123",
      "AUCIFOS", "AUC Infinity Obs Norm by SA", "AUCINF_obs_S", "Plasma/Blood/Serum", "124",
      "AUCIFOUB", "AUC Infinity Obs, Unbound Drug", "AUCIFOUB", "Plasma/Blood/Serum", "125",
      "AUCIFOW", "AUC Infinity Obs Norm by WT", "AUCINF_obs_W", "Plasma/Blood/Serum", "126",
      "AUCIFPB", "AUC Infinity Pred Norm by BMI", "AUCINF_pred_B", "Plasma/Blood/Serum", "127",
      "AUCIFPD", "AUC Infinity Pred Norm by Dose", "AUCINF_pred_D", "Plasma/Blood/Serum", "128",
      "AUCIFPS", "AUC Infinity Pred Norm by SA", "AUCINF_pred_S", "Plasma/Blood/Serum", "129",
      "AUCIFPUB", "AUC Infinity Pred, Unbound Drug", "AUCIFPUB", "Plasma/Blood/Serum", "130",
      "AUCIFPW", "AUC Infinity Pred Norm by WT", "AUCINF_pred_W", "Plasma/Blood/Serum", "131",
      "AUCINTB", "AUC from T1 to T2 Norm by BMI", "AUC_B_T1_T2_UNIT", "Plasma/Blood/Serum", "132",
      "AUCINTD", "AUC from T1 to T2 Norm by Dose", "AUC_D_T1_T2_UNIT", "Plasma/Blood/Serum", "133",
      "AUCINTS", "AUC from T1 to T2 Norm by SA", "AUC_S_T1_T2_UNIT", "Plasma/Blood/Serum", "134",
      "AUCINTW", "AUC from T1 to T2 Norm by WT", "AUC_W_T1_T2_UNIT", "Plasma/Blood/Serum", "135",
      "AUCLSTB", "AUC to Last Nonzero Conc Norm by BMI", "AUClast_B", "Plasma/Blood/Serum", "136",
      "AUCLSTD", "AUC to Last Nonzero Conc Norm by Dose", "AUClast_D", "Plasma/Blood/Serum", "137",
      "AUCLSTLN", "AUC to Last Nonzero Conc LN Transformed", "AUCLSTLN", "Plasma/Blood/Serum", "138",
      "AUCLSTS", "AUC to Last Nonzero Conc Norm by SA", "AUClast_S", "Plasma/Blood/Serum", "139",
      "AUCLSTUB", "AUC to Last Nonzero Conc, Unbound Drug", "AUCLSTUB", "Plasma/Blood/Serum", "140",
      "AUCLSTW", "AUC to Last Nonzero Conc Norm by WT", "AUClast_W", "Plasma/Blood/Serum", "141",
      "AUCTAUB", "AUC Over Dosing Interval Norm by BMI", "AUC_TAU_B", "Plasma/Blood/Serum", "142",
      "AUCTAUD", "AUC Over Dosing Interval Norm by Dose", "AUC_TAU_D", "Plasma/Blood/Serum", "143",
      "AUCTAUS", "AUC Over Dosing Interval Norm by SA", "AUC_TAU_S", "Plasma/Blood/Serum", "144",
      "AUCTAUW", "AUC Over Dosing Interval Norm by WT", "AUC_TAU_W", "Plasma/Blood/Serum", "145",
      "AUMCIFOB", "AUMC Infinity Obs Norm by BMI", "AUMCINF_obs_B", "Plasma/Blood/Serum", "146",
      "AUMCIFOD", "AUMC Infinity Obs Norm by Dose", "AUMCINF_obs_D", "Plasma/Blood/Serum", "147",
      "AUMCIFOS", "AUMC Infinity Obs Norm by SA", "AUMCINF_obs_S", "Plasma/Blood/Serum", "148",
      "AUMCIFOW", "AUMC Infinity Obs Norm by WT", "AUMCINF_obs_W", "Plasma/Blood/Serum", "149",
      "AUMCIFPB", "AUMC Infinity Pred Norm by BMI", "AUMCINF_pred_B", "Plasma/Blood/Serum", "150",
      "AUMCIFPD", "AUMC Infinity Pred Norm by Dose", "AUMCINF_pred_D", "Plasma/Blood/Serum", "151",
      "AUMCIFPS", "AUMC Infinity Pred Norm by SA", "AUMCINF_pred_S", "Plasma/Blood/Serum", "152",
      "AUMCIFPW", "AUMC Infinity Pred Norm by WT", "AUMCINF_pred_W", "Plasma/Blood/Serum", "153",
      "AUMCLSTB", "AUMC to Last Nonzero Conc Norm by BMI", "AUMClast_B", "Plasma/Blood/Serum", "154",
      "AUMCLSTD", "AUMC to Last Nonzero Conc Norm by Dose", "AUMClast_D", "Plasma/Blood/Serum", "155",
      "AUMCLSTS", "AUMC to Last Nonzero Conc Norm by SA", "AUMClast_S", "Plasma/Blood/Serum", "156",
      "AUMCLSTW", "AUMC to Last Nonzero Conc Norm by WT", "AUMClast_W", "Plasma/Blood/Serum", "157",
      "AUMCTAUB", "AUMC Over Dosing Interval Norm by BMI", "AUMCTAUB", "Plasma/Blood/Serum", "158",
      "AUMCTAUD", "AUMC Over Dosing Interval Norm by Dose", "AUMCTAUD", "Plasma/Blood/Serum", "159",
      "AUMCTAUS", "AUMC Over Dosing Interval Norm by SA", "AUMCTAUS", "Plasma/Blood/Serum", "160",
      "AUMCTAUW", "AUMC Over Dosing Interval Norm by WT", "AUMCTAUW", "Plasma/Blood/Serum", "161",
      "AURCALLB", "AURC All Norm by BMI", "AURCALLB", "Plasma/Blood/Serum", "162",
      "AURCALLD", "AURC All Norm by Dose", "AURCALLD", "Plasma/Blood/Serum", "163",
      "AURCALLS", "AURC All Norm by SA", "AURCALLS", "Plasma/Blood/Serum", "164",
      "AURCALLW", "AURC All Norm by WT", "AURCALLW", "Plasma/Blood/Serum", "165",
      "AURCIFOB", "AURC Infinity Obs Norm by BMI", "AURCIFOB", "Plasma/Blood/Serum", "166",
      "AURCIFOD", "AURC Infinity Obs Norm by Dose", "AURCIFOD", "Plasma/Blood/Serum", "167",
      "AURCIFOS", "AURC Infinity Obs Norm by SA", "AURCIFOS", "Plasma/Blood/Serum", "168",
      "AURCIFOW", "AURC Infinity Obs Norm by WT", "AURCIFOW", "Plasma/Blood/Serum", "169",
      "AURCIFPB", "AURC Infinity Pred Norm by BMI", "AURCIFPB", "Plasma/Blood/Serum", "170",
      "AURCIFPD", "AURC Infinity Pred Norm by Dose", "AURCIFPD", "Plasma/Blood/Serum", "171",
      "AURCIFPS", "AURC Infinity Pred Norm by SA", "AURCIFPS", "Plasma/Blood/Serum", "172",
      "AURCIFPW", "AURC Infinity Pred Norm by WT", "AURCIFPW", "Plasma/Blood/Serum", "173",
      "AURCINT", "AURC from T1 to T2", "AURCINT_T1_T2_UNIT", "Plasma/Blood/Serum", "174",
      "AURCINTB", "AURC from T1 to T2 Norm by BMI", "AURCINTB_T1_T2_UNIT", "Plasma/Blood/Serum", "175",
      "AURCINTD", "AURC from T1 to T2 Norm by Dose", "AURCINTD_T1_T2_UNIT", "Plasma/Blood/Serum", "176",
      "AURCINTS", "AURC from T1 to T2 Norm by SA", "AURCINTS_T1_T2_UNIT", "Plasma/Blood/Serum", "177",
      "AURCINTW", "AURC from T1 to T2 Norm by WT", "AURCINTW_T1_T2_UNIT", "Plasma/Blood/Serum", "178",
      "AURCLSTB", "AURC to Last Nonzero Rate Norm by BMI", "AURCLSTB", "Plasma/Blood/Serum", "179",
      "AURCLSTD", "AURC to Last Nonzero Rate Norm by Dose", "AURCLSTD", "Plasma/Blood/Serum", "180",
      "AURCLSTS", "AURC to Last Nonzero Rate Norm by SA", "AURCLSTS", "Plasma/Blood/Serum", "181",
      "AURCLSTW", "AURC to Last Nonzero Rate Norm by WT", "AURCLSTW", "Plasma/Blood/Serum", "182",
      "C0B", "Initial Conc Norm by BMI", "C0B", "Plasma/Blood/Serum", "183",
      "C0D", "Initial Conc Norm by Dose", "C0D", "Plasma/Blood/Serum", "184",
      "C0S", "Initial Conc Norm by SA", "C0S", "Plasma/Blood/Serum", "185",
      "C0W", "Initial Conc Norm by WT", "C0W", "Plasma/Blood/Serum", "186",
      "CAVGB", "Average Conc Norm by BMI", "CAVGB", "Plasma/Blood/Serum", "187",
      "CAVGD", "Average Conc Norm by Dose", "CAVGD", "Plasma/Blood/Serum", "188",
      "CAVGINT", "Average Conc from T1 to T2", "CAVGINT_T1_T2_UNIT", "Plasma/Blood/Serum", "189",
      "CAVGINTB", "Average Conc from T1 to T2 Norm by BMI", "CAVGINTB_T1_T2_UNIT", "Plasma/Blood/Serum", "190",
      "CAVGINTD", "Average Conc from T1 to T2 Norm by Dose", "CAVGINTD_T1_T2_UNIT", "Plasma/Blood/Serum", "191",
      "CAVGINTS", "Average Conc from T1 to T2 Norm by SA", "CAVGINTS_T1_T2_UNIT", "Plasma/Blood/Serum", "192",
      "CAVGINTW", "Average Conc from T1 to T2 Norm by WT", "CAVGINTW_T1_T2_UNIT", "Plasma/Blood/Serum", "193",
      "CAVGS", "Average Conc Norm by SA", "CAVGS", "Plasma/Blood/Serum", "194",
      "CAVGW", "Average Conc Norm by WT", "CAVGW", "Plasma/Blood/Serum", "195",
      "CHTMAX", "Concentration at Half Tmax", "CHTMAX", "Plasma/Blood/Serum", "196",
      "CLFOB", "Total CL Obs by F Norm by BMI", "CLFOB", "Plasma/Blood/Serum", "197",
      "CLFOD", "Total CL Obs by F Norm by Dose", "CLFOD", "Plasma/Blood/Serum", "198",
      "CLFOS", "Total CL Obs by F Norm by SA", "CLFOS", "Plasma/Blood/Serum", "199",
      "CLFOW", "Total CL Obs by F Norm by WT", "CLFOW", "Plasma/Blood/Serum", "200",
      "CLFPB", "Total CL Pred by F Norm by BMI", "CLFPB", "Plasma/Blood/Serum", "201",
      "CLFPD", "Total CL Pred by F Norm by Dose", "CLFPD", "Plasma/Blood/Serum", "202",
      "CLFPS", "Total CL Pred by F Norm by SA", "CLFPS", "Plasma/Blood/Serum", "203",
      "CLFPW", "Total CL Pred by F Norm by WT", "CLFPW", "Plasma/Blood/Serum", "204",
      "CLFTAU", "Total CL by F for Dose Int", "CLFTAU", "Plasma/Blood/Serum", "205",
      "CLFTAUB", "Total CL by F for Dose Int Norm by BMI", "CLFTAUB", "Plasma/Blood/Serum", "206",
      "CLFTAUD", "Total CL by F for Dose Int Norm by Dose", "CLFTAUD", "Plasma/Blood/Serum", "207",
      "CLFTAUS", "Total CL by F for Dose Int Norm by SA", "CLFTAUS", "Plasma/Blood/Serum", "208",
      "CLFTAUW", "Total CL by F for Dose Int Norm by WT", "CLFTAUW", "Plasma/Blood/Serum", "209",
      "CLFUB", "Apparent CL for Unbound Drug", "CLFUB", "Plasma/Blood/Serum", "210",
      "CLOB", "Total CL Obs Norm by BMI", "CLOB", "Plasma/Blood/Serum", "211",
      "CLOD", "Total CL Obs Norm by Dose", "CLOD", "Plasma/Blood/Serum", "212",
      "CLOS", "Total CL Obs Norm by SA", "CLOS", "Plasma/Blood/Serum", "213",
      "CLOUB", "Total CL Obs for Unbound Drug", "CLOUB", "Plasma/Blood/Serum", "214",
      "CLOW", "Total CL Obs Norm by WT", "CLOW", "Plasma/Blood/Serum", "215",
      "CLPB", "Total CL Pred Norm by BMI", "CLPB", "Plasma/Blood/Serum", "216",
      "CLPD", "Total CL Pred Norm by Dose", "CLPD", "Plasma/Blood/Serum", "217",
      "CLPS", "Total CL Pred Norm by SA", "CLPS", "Plasma/Blood/Serum", "218",
      "CLPUB", "Total CL Pred for Unbound Drug", "CLPUB", "Plasma/Blood/Serum", "219",
      "CLPW", "Total CL Pred Norm by WT", "CLPW", "Plasma/Blood/Serum", "220",
      "CLRPCLEV", "Renal CL as Pct CL EV", "CLRPCLEV", "Urine", "221",
      "CLRPCLIV", "Renal CL as Pct CL IV", "CLRPCLIV", "Urine", "222",
      "CLSTB", "Last Nonzero Conc Norm by BMI", "CLSTB", "Plasma/Blood/Serum", "223",
      "CLSTD", "Last Nonzero Conc Norm by Dose", "CLSTD", "Plasma/Blood/Serum", "224",
      "CLSTS", "Last Nonzero Conc Norm by SA", "CLSTS", "Plasma/Blood/Serum", "225",
      "CLSTW", "Last Nonzero Conc Norm by WT", "CLSTW", "Plasma/Blood/Serum", "226",
      "CLTAU", "Total CL for Dose Int", "CLTAU", "Plasma/Blood/Serum", "227",
      "CLTAUB", "Total CL for Dose Int Norm by BMI", "CLTAUB", "Plasma/Blood/Serum", "228",
      "CLTAUD", "Total CL for Dose Int Norm by Dose", "CLTAUD", "Plasma/Blood/Serum", "229",
      "CLTAUS", "Total CL for Dose Int Norm by SA", "CLTAUS", "Plasma/Blood/Serum", "230",
      "CLTAUW", "Total CL for Dose Int Norm by WT", "CLTAUW", "Plasma/Blood/Serum", "231",
      "CMAXB", "Max Conc Norm by BMI", "CMAX_B", "Plasma/Blood/Serum", "232",
      "CMAXLN", "Max Conc LN Transformed", "CMAXLN", "Plasma/Blood/Serum", "233",
      "CMAXS", "Max Conc Norm by SA", "CMAXS", "Plasma/Blood/Serum", "234",
      "CMAXUB", "Max Conc, Unbound Drug", "CMAXUB", "Plasma/Blood/Serum", "235",
      "CMAXW", "Max Conc Norm by WT", "CMAXW", "Plasma/Blood/Serum", "236",
      "CMINB", "Min Conc Norm by BMI", "CMINB", "Plasma/Blood/Serum", "237",
      "CMIND", "Min Conc Norm by Dose", "CMIND", "Plasma/Blood/Serum", "238",
      "CMINS", "Min Conc Norm by SA", "CMINS", "Plasma/Blood/Serum", "239",
      "CMINW", "Min Conc Norm by WT", "CMINW", "Plasma/Blood/Serum", "240",
      "CONC", "Concentration", "CONC", "Plasma/Blood/Serum", "241",
      "CONCB", "Conc by BMI", "CONCB", "Plasma/Blood/Serum", "242",
      "CONCD", "Conc by Dose", "CONCD", "Plasma/Blood/Serum", "243",
      "CONCS", "Conc by SA", "CONCS", "Plasma/Blood/Serum", "244",
      "CONCW", "Conc by WT", "CONCW", "Plasma/Blood/Serum", "245",
      "CTROUGH", "Conc Trough", "CTROUGH", "Plasma/Blood/Serum", "246",
      "CTROUGHB", "Conc Trough by BMI", "CTROUGHB", "Plasma/Blood/Serum", "247",
      "CTROUGHD", "Conc Trough by Dose", "CTROUGHD", "Plasma/Blood/Serum", "248",
      "CTROUGHS", "Conc Trough by SA", "CTROUGHS", "Plasma/Blood/Serum", "249",
      "CTROUGHW", "Conc Trough by WT", "CTROUGHW", "Plasma/Blood/Serum", "250",
      "EFFHL", "Effective Half-Life", "EFFHL", "Plasma/Blood/Serum", "251",
      "ERINT", "Excret Rate from T1 to T2", "ERINT_T1_T2_UNIT", "Plasma/Blood/Serum", "252",
      "ERINTB", "Excret Rate from T1 to T2 Norm by BMI", "ERINTB_T1_T2_UNIT", "Plasma/Blood/Serum", "253",
      "ERINTD", "Excret Rate from T1 to T2 Norm by Dose", "ERINTD_T1_T2_UNIT", "Plasma/Blood/Serum", "254",
      "ERINTS", "Excret Rate from T1 to T2 Norm by SA", "ERINTS_T1_T2_UNIT", "Plasma/Blood/Serum", "255",
      "ERINTW", "Excret Rate from T1 to T2 Norm by WT", "ERINTW_T1_T2_UNIT", "Plasma/Blood/Serum", "256",
      "ERLSTB", "Last Meas Excretion Rate Norm by BMI", "ERLSTB", "Plasma/Blood/Serum", "257",
      "ERLSTD", "Last Meas Excretion Rate Norm by Dose", "ERLSTD", "Plasma/Blood/Serum", "258",
      "ERLSTS", "Last Meas Excretion Rate Norm by SA", "ERLSTS", "Plasma/Blood/Serum", "259",
      "ERLSTW", "Last Meas Excretion Rate Norm by WT", "ERLSTW", "Plasma/Blood/Serum", "260",
      "ERMAXB", "Max Excretion Rate Norm by BMI", "ERMAXB", "Plasma/Blood/Serum", "261",
      "ERMAXD", "Max Excretion Rate Norm by Dose", "ERMAXD", "Plasma/Blood/Serum", "262",
      "ERMAXS", "Max Excretion Rate Norm by SA", "ERMAXS", "Plasma/Blood/Serum", "263",
      "ERMAXW", "Max Excretion Rate Norm by WT", "ERMAXW", "Plasma/Blood/Serum", "264",
      "ERTLST", "Midpoint of Interval of Last Nonzero ER", "ERTLST", "Plasma/Blood/Serum", "265",
      "FABS", "Absolute Bioavailability", "FABS", "Plasma/Blood/Serum", "266",
      "FB", "Fraction Bound", "FB", "Plasma/Blood/Serum", "267",
      "FREL", "Relative Bioavailability", "FREL", "Plasma/Blood/Serum", "268",
      "FREXINT", "Fract Excr from T1 to T2", "FREXINT_T1_T2_UNIT", "Plasma/Blood/Serum", "269",
      "FU", "Fraction Unbound", "FU", "Plasma/Blood/Serum", "270",
      "HDCL", "Hemodialysis Clearance", "HDCL", "Plasma/Blood/Serum", "271",
      "HDER", "Hemodialysis Extraction Ratio", "HDER", "Plasma/Blood/Serum", "272",
      "HTMAX", "Half Tmax", "HTMAX", "Plasma/Blood/Serum", "273",
      "LAMZLTAU", "Lambda z Lower Limit TAU", "LAMZLTAU", "Plasma/Blood/Serum", "274",
      "LAMZNTAU", "Number of Points for Lambda z TAU", "LAMZNTAU", "Plasma/Blood/Serum", "275",
      "LAMZSPN", "Lambda z Span", "LAMZSPN", "Plasma/Blood/Serum", "276",
      "LAMZTAU", "Lambda z TAU", "LAMZTAU", "Plasma/Blood/Serum", "277",
      "LAMZUTAU", "Lambda z Upper Limit TAU", "LAMZUTAU", "Plasma/Blood/Serum", "278",
      "MAT", "Mean Absorption Time", "MAT", "Plasma/Blood/Serum", "279",
      "MRAUCIFO", "Metabolite Ratio for AUC Infinity Obs", "MRAUCIFO", "Plasma/Blood/Serum", "280",
      "MRAUCIFP", "Metabolite Ratio for AUC Infinity Pred", "MRAUCIFP", "Plasma/Blood/Serum", "281",
      "MRAUCINT", "Metabolite Ratio AUC from T1 to T2", "MRAUCINT_T1_T2_UNIT", "Plasma/Blood/Serum", "282",
      "MRAUCLST", "Metabolite Ratio AUC Last Nonzero Conc", "MRAUCLST", "Plasma/Blood/Serum", "283",
      "MRAUCTAU", "Metabolite Ratio for AUC Dosing Interval", "MRAUCTAU", "Plasma/Blood/Serum", "284",
      "MRCMAX", "Metabolite Ratio for Max Conc", "MRCMAX", "Plasma/Blood/Serum", "285",
      "MRTEVIFO", "MRT Extravasc Infinity Obs", "MRTEVIFO", "Plasma/Blood/Serum", "286",
      "MRTEVIFP", "MRT Extravasc Infinity Pred", "MRTEVIFP", "Plasma/Blood/Serum", "287",
      "MRTEVLST", "MRT Extravasc to Last Nonzero Conc", "MRTEVLST", "Plasma/Blood/Serum", "288",
      "MRTIVIFO", "MRT Intravasc Infinity Obs", "MRTIVIFO", "Plasma/Blood/Serum", "289",
      "MRTIVIFP", "MRT Intravasc Infinity Pred", "MRTIVIFP", "Plasma/Blood/Serum", "290",
      "MRTIVLST", "MRT Intravasc to Last Nonzero Conc", "MRTIVLST", "Plasma/Blood/Serum", "291",
      "NRENALCL", "Nonrenal CL", "NRENALCL", "Urine", "292",
      "NRENLCLB", "Nonrenal CL Norm by BMI", "NRENLCLB", "Urine", "293",
      "NRENLCLD", "Nonrenal CL Norm by Dose", "NRENLCLD", "Urine", "294",
      "NRENLCLS", "Nonrenal CL Norm by SA", "NRENLCLS", "Urine", "295",
      "NRENLCLW", "Nonrenal CL Norm by WT", "NRENLCLW", "Urine", "296",
      "PTROUGHR", "Peak Trough Ratio", "PTROUGHR", "Plasma/Blood/Serum", "297",
      "RAAUC", "Ratio AUC", "RAAUC", "Plasma/Blood/Serum", "298",
      "RAAUCIFO", "Ratio AUC Infinity Obs", "RAAUCIFO", "Plasma/Blood/Serum", "299",
      "RAAUCIFP", "Ratio AUC Infinity Pred", "RAAUCIFP", "Plasma/Blood/Serum", "300",
      "RACMAX", "Ratio CMAX", "RACMAX", "Plasma/Blood/Serum", "301",
      "RAMAXMIN", "Ratio of CMAX to CMIN", "RAMAXMIN", "Plasma/Blood/Serum", "302",
      "RCAMIFO", "Amt Rec Infinity Obs", "RCAMIFO", "Plasma/Blood/Serum", "303",
      "RCAMIFOB", "Amt Rec Infinity Obs Norm by BMI", "RCAMIFOB", "Plasma/Blood/Serum", "304",
      "RCAMIFOS", "Amt Rec Infinity Obs Norm by SA", "RCAMIFOS", "Plasma/Blood/Serum", "305",
      "RCAMIFOW", "Amt Rec Infinity Obs Norm by WT", "RCAMIFOW", "Plasma/Blood/Serum", "306",
      "RCAMIFP", "Amt Rec Infinity Pred", "RCAMIFP", "Plasma/Blood/Serum", "307",
      "RCAMIFPB", "Amt Rec Infinity Pred Norm by BMI", "RCAMIFPB", "Plasma/Blood/Serum", "308",
      "RCAMIFPS", "Amt Rec Infinity Pred Norm by SA", "RCAMIFPS", "Plasma/Blood/Serum", "309",
      "RCAMIFPW", "Amt Rec Infinity Pred Norm by WT", "RCAMIFPW", "Plasma/Blood/Serum", "310",
      "RCAMINTB", "Amt Rec from T1 to T2 Norm by BMI", "RCAMINTB_T1_T2_UNIT", "Plasma/Blood/Serum", "311",
      "RCAMINTS", "Amt Rec from T1 to T2 Norm by SA", "RCAMINTS_T1_T2_UNIT", "Plasma/Blood/Serum", "312",
      "RCAMINTW", "Amt Rec from T1 to T2 Norm by WT", "RCAMINTW_T1_T2_UNIT", "Plasma/Blood/Serum", "313",
      "RCAMTAU", "Amt Rec Over Dosing Interval", "RCAMTAU", "Plasma/Blood/Serum", "314",
      "RCAMTAUB", "Amt Rec Over Dosing Interval Norm by BMI", "RCAMTAUB", "Plasma/Blood/Serum", "315",
      "RCAMTAUS", "Amt Rec Over Dosing Interval Norm by SA", "RCAMTAUS", "Plasma/Blood/Serum", "316",
      "RCAMTAUW", "Amt Rec Over Dosing Interval Norm by WT", "RCAMTAUW", "Plasma/Blood/Serum", "317",
      "RCPCIFO", "Pct Rec Infinity Obs", "RCPCIFO", "Plasma/Blood/Serum", "318",
      "RCPCIFOB", "Pct Rec Infinity Obs Norm by BMI", "RCPCIFOB", "Plasma/Blood/Serum", "319",
      "RCPCIFOS", "Pct Rec Infinity Obs Norm by SA", "RCPCIFOS", "Plasma/Blood/Serum", "320",
      "RCPCIFOW", "Pct Rec Infinity Obs Norm by WT", "RCPCIFOW", "Plasma/Blood/Serum", "321",
      "RCPCIFP", "Pct Rec Infinity Pred", "RCPCIFP", "Plasma/Blood/Serum", "322",
      "RCPCIFPB", "Pct Rec Infinity Pred Norm by BMI", "RCPCIFPB", "Plasma/Blood/Serum", "323",
      "RCPCIFPS", "Pct Rec Infinity Pred Norm by SA", "RCPCIFPS", "Plasma/Blood/Serum", "324",
      "RCPCIFPW", "Pct Rec Infinity Pred Norm by WT", "RCPCIFPW", "Plasma/Blood/Serum", "325",
      "RCPCINTB", "Pct Rec from T1 to T2 Norm by BMI", "RCPCINTB_T1_T2_UNIT", "Plasma/Blood/Serum", "326",
      "RCPCINTS", "Pct Rec from T1 to T2 Norm by SA", "RCPCINTS_T1_T2_UNIT", "Plasma/Blood/Serum", "327",
      "RCPCINTW", "Pct Rec from T1 to T2 Norm by WT", "RCPCINTW_T1_T2_UNIT", "Plasma/Blood/Serum", "328",
      "RCPCLST", "Pct Rec to Last Nonzero Conc", "RCPCLST", "Plasma/Blood/Serum", "329",
      "RCPCTAU", "Pct Rec Over Dosing Interval", "RCPCTAU", "Plasma/Blood/Serum", "330",
      "RCPCTAUB", "Pct Rec Over Dosing Interval Norm by BMI", "RCPCTAUB", "Plasma/Blood/Serum", "331",
      "RCPCTAUS", "Pct Rec Over Dosing Interval Norm by SA", "RCPCTAUS", "Plasma/Blood/Serum", "332",
      "RCPCTAUW", "Pct Rec Over Dosing Interval Norm by WT", "RCPCTAUW", "Plasma/Blood/Serum", "333",
      "RENALCLB", "Renal CL Norm by BMI", "RENALCLB", "Urine", "334",
      "RENALCLD", "Renal CL Norm by Dose", "RENALCLD", "Urine", "335",
      "RENALCLS", "Renal CL Norm by SA", "RENALCLS", "Urine", "336",
      "RENALCLW", "Renal CL Norm by WT", "RENALCLW", "Urine", "337",
      "RENCLTAU", "Renal CL for Dose Int", "RENCLTAU", "Urine", "338",
      "RNCLINT", "Renal CL from T1 to T2", "RNCLINT_T1_T2_UNIT", "Urine", "339",
      "RNCLINTB", "Renal CL from T1 to T2 Norm by BMI", "RNCLINTB_T1_T2_UNIT", "Urine", "340",
      "RNCLINTD", "Renal CL from T1 to T2 Norm by Dose", "RNCLINTD_T1_T2_UNIT", "Urine", "341",
      "RNCLINTS", "Renal CL from T1 to T2 Norm by SA", "RNCLINTS_T1_T2_UNIT", "Urine", "342",
      "RNCLINTW", "Renal CL from T1 to T2 Norm by WT", "RNCLINTW_T1_T2_UNIT", "Urine", "343",
      "RNCLTAUB", "Renal CL for Dose Int Norm by BMI", "RNCLTAUB", "Urine", "344",
      "RNCLTAUD", "Renal CL for Dose Int Norm by Dose", "RNCLTAUD", "Urine", "345",
      "RNCLTAUS", "Renal CL for Dose Int Norm by SA", "RNCLTAUS", "Urine", "346",
      "RNCLTAUW", "Renal CL for Dose Int Norm by WT", "RNCLTAUW", "Urine", "347",
      "RNCLUB", "Renal CL for Unbound Drug", "RNCLUB", "Urine", "348",
      "SRAUC", "Stationarity Ratio AUC", "SRAUC", "Plasma/Blood/Serum", "349",
      "SWING", "Swing", "SWING", "Plasma/Blood/Serum", "350",
      "TAUHL", "Half-Life TAU", "TAUHL", "Plasma/Blood/Serum", "351",
      "TBBL", "Time Below Baseline", "Time_Below_B", "Plasma/Blood/Serum", "352",
      "TROUGHPR", "Trough Peak Ratio", "TROUGHPR", "Plasma/Blood/Serum", "353",
      "V0", "Vol Dist Initial", "V0", "Plasma/Blood/Serum", "354",
      "V0B", "Vol Dist Initial Norm by BMI", "V0B", "Plasma/Blood/Serum", "355",
      "V0D", "Vol Dist Initial Norm by Dose", "V0D", "Plasma/Blood/Serum", "356",
      "V0S", "Vol Dist Initial Norm by SA", "V0S", "Plasma/Blood/Serum", "357",
      "V0W", "Vol Dist Initial Norm by WT", "V0W", "Plasma/Blood/Serum", "358",
      "VSSOB", "Vol Dist Steady State Obs Norm by BMI", "VSSOB", "Plasma/Blood/Serum", "359",
      "VSSOBD", "Vol Dist Steady State Obs by B", "VSSOBD", "Plasma/Blood/Serum", "360",
      "VSSOD", "Vol Dist Steady State Obs Norm by Dose", "VSSOD", "Plasma/Blood/Serum", "361",
      "VSSOF", "Vol Dist Steady State Obs by F", "VSSOF", "Plasma/Blood/Serum", "362",
      "VSSOS", "Vol Dist Steady State Obs Norm by SA", "VSSOS", "Plasma/Blood/Serum", "363",
      "VSSOUB", "Vol Dist Steady State Obs by UB", "VSSOUB", "Plasma/Blood/Serum", "364",
      "VSSOW", "Vol Dist Steady State Obs Norm by WT", "VSSOW", "Plasma/Blood/Serum", "365",
      "VSSPB", "Vol Dist Steady State Pred Norm by BMI", "VSSPB", "Plasma/Blood/Serum", "366",
      "VSSPBD", "Vol Dist Steady State Pred by B", "VSSPBD", "Plasma/Blood/Serum", "367",
      "VSSPD", "Vol Dist Steady State Pred Norm by Dose", "VSSPD", "Plasma/Blood/Serum", "368",
      "VSSPF", "Vol Dist Steady State Pred by F", "VSSPF", "Plasma/Blood/Serum", "369",
      "VSSPS", "Vol Dist Steady State Pred Norm by SA", "VSSPS", "Plasma/Blood/Serum", "370",
      "VSSPUB", "Vol Dist Steady State Pred by UB", "VSSPUB", "Plasma/Blood/Serum", "371",
      "VSSPW", "Vol Dist Steady State Pred Norm by WT", "VSSPW", "Plasma/Blood/Serum", "372",
      "VZ", "Vol Z", "Vz", "Plasma/Blood/Serum", "373",
      "VZF", "Vol Z by F", "Vz_F", "Plasma/Blood/Serum", "374",
      "VZFOB", "Vz Obs by F Norm by BMI", "VZFOB", "Plasma/Blood/Serum", "375",
      "VZFOD", "Vz Obs by F Norm by Dose", "VZFOD", "Plasma/Blood/Serum", "376",
      "VZFOS", "Vz Obs by F Norm by SA", "VZFOS", "Plasma/Blood/Serum", "377",
      "VZFOUB", "Vz Obs by F for UB", "VZFOUB", "Plasma/Blood/Serum", "378",
      "VZFOW", "Vz Obs by F Norm by WT", "VZFOW", "Plasma/Blood/Serum", "379",
      "VZFPB", "Vz Pred by F Norm by BMI", "VZFPB", "Plasma/Blood/Serum", "380",
      "VZFPD", "Vz Pred by F Norm by Dose", "VZFPD", "Plasma/Blood/Serum", "381",
      "VZFPS", "Vz Pred by F Norm by SA", "VZFPS", "Plasma/Blood/Serum", "382",
      "VZFPUB", "Vz Pred by F for UB", "VZFPUB", "Plasma/Blood/Serum", "383",
      "VZFPW", "Vz Pred by F Norm by WT", "VZFPW", "Plasma/Blood/Serum", "384",
      "VZFTAU", "Vz for Dose Int by F", "VZFTAU", "Plasma/Blood/Serum", "385",
      "VZFTAUB", "Vz for Dose Int by F Norm by BMI", "VZFTAUB", "Plasma/Blood/Serum", "386",
      "VZFTAUD", "Vz for Dose Int by F Norm by Dose", "VZFTAUD", "Plasma/Blood/Serum", "387",
      "VZFTAUS", "Vz for Dose Int by F Norm by SA", "VZFTAUS", "Plasma/Blood/Serum", "388",
      "VZFTAUW", "Vz for Dose Int by F Norm by WT", "VZFTAUW", "Plasma/Blood/Serum", "389",
      "VZOB", "Vz Obs Norm by BMI", "VZOB", "Plasma/Blood/Serum", "390",
      "VZOD", "Vz Obs Norm by Dose", "VZOD", "Plasma/Blood/Serum", "391",
      "VZOS", "Vz Obs Norm by SA", "VZOS", "Plasma/Blood/Serum", "392",
      "VZOUB", "Vz Obs for UB", "VZOUB", "Plasma/Blood/Serum", "393",
      "VZOW", "Vz Obs Norm by WT", "VZOW", "Plasma/Blood/Serum", "394",
      "VZPB", "Vz Pred Norm by BMI", "VZPB", "Plasma/Blood/Serum", "395",
      "VZPD", "Vz Pred Norm by Dose", "VZPD", "Plasma/Blood/Serum", "396",
      "VZPS", "Vz Pred Norm by SA", "VZPS", "Plasma/Blood/Serum", "397",
      "VZPUB", "Vz Pred for UB", "VZPUB", "Plasma/Blood/Serum", "398"
    ),
    ncol = 5,
    byrow = TRUE
  ))
  colnames(pk_dataset) <- c("PARAMCD", "PARAM", "TLG_DISPLAY", "MATRIX", "TLG_ORDER")
  pk_dataset
}

1		#' Count patients with toxicity grades that have worsened from baseline by highest grade post-baseline
2		#'
3		#' @description `r lifecycle::badge("stable")`
4		#'
5		#' The analyze function [count_abnormal_lab_worsen_by_baseline()] creates a layout element to count patients with
6		#' analysis toxicity grades which have worsened from baseline, categorized by highest (worst) grade post-baseline.
7		#'
8		#' This function analyzes primary analysis variable `var` which indicates analysis toxicity grades. Additional
9		#' analysis variables that can be supplied as a list via the `variables` parameter are `id` (defaults to `USUBJID`),
10		#' a variable to indicate unique subject identifiers, `baseline_var` (defaults to `BTOXGR`), a variable to indicate
11		#' baseline toxicity grades, and `direction_var` (defaults to `GRADDIR`), a variable to indicate toxicity grade
12		#' directions of interest to include (e.g. `"H"` (high), `"L"` (low), or `"B"` (both)).
13		#'
14		#' For the direction(s) specified in `direction_var`, patient counts by worst grade for patients who have
15		#' worsened from baseline are calculated as follows:
16		#' * `1` to `4`: The number of patients who have worsened from their baseline grades with worst
17		#' grades 1-4, respectively.
18		#' * `Any`: The total number of patients who have worsened from their baseline grades.
19		#'
20		#' Fractions are calculated by dividing the above counts by the number of patients who's analysis toxicity grades
21		#' have worsened from baseline toxicity grades during treatment.
22		#'
23		#' Prior to using this function in your table layout you must use [rtables::split_rows_by()] to create a row
24		#' split on variable `direction_var`.
25		#'
26		#' @inheritParams argument_convention
27		#' @param variables (named `list` of `string`)\cr list of additional analysis variables including:
28		#' * `id` (`string`)\cr subject variable name.
29		#' * `baseline_var` (`string`)\cr name of the data column containing baseline toxicity variable.
30		#' * `direction_var` (`string`)\cr see `direction_var` for more details.
31		#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("abnormal_by_worst_grade_worsen")`
32		#' to see all available statistics.
33		#'
34		#' @seealso Relevant helper functions [h_adlb_worsen()] and [h_worsen_counter()] which are used within
35		#' [s_count_abnormal_lab_worsen_by_baseline()] to process input data.
36		#'
37		#' @name abnormal_by_worst_grade_worsen
38		#' @order 1
39		NULL
40
41		#' Helper function to prepare ADLB with worst labs
42		#'
43		#' @description `r lifecycle::badge("stable")`
44		#'
45		#' Helper function to prepare a `df` for generate the patient count shift table.
46		#'
47		#' @param adlb (`data.frame`)\cr ADLB data frame.
48		#' @param worst_flag_low (named `vector`)\cr worst low post-baseline lab grade flag variable. See how this is
49		#' implemented in the following examples.
50		#' @param worst_flag_high (named `vector`)\cr worst high post-baseline lab grade flag variable. See how this is
51		#' implemented in the following examples.
52		#' @param direction_var (`string`)\cr name of the direction variable specifying the direction of the shift table of
53		#' interest. Only lab records flagged by `L`, `H` or `B` are included in the shift table.
54		#' * `L`: low direction only
55		#' * `H`: high direction only
56		#' * `B`: both low and high directions
57		#'
58		#' @return `h_adlb_worsen()` returns the `adlb` `data.frame` containing only the
59		#' worst labs specified according to `worst_flag_low` or `worst_flag_high` for the
60		#' direction specified according to `direction_var`. For instance, for a lab that is
61		#' needed for the low direction only, only records flagged by `worst_flag_low` are
62		#' selected. For a lab that is needed for both low and high directions, the worst
63		#' low records are selected for the low direction, and the worst high record are selected
64		#' for the high direction.
65		#'
66		#' @seealso [abnormal_by_worst_grade_worsen]
67		#'
68		#' @examples
69		#' library(dplyr)
70		#'
71		#' # The direction variable, GRADDR, is based on metadata
72		#' adlb <- tern_ex_adlb %>%
73		#' mutate(
74		#' GRADDR = case_when(
75		#' PARAMCD == "ALT" ~ "B",
76		#' PARAMCD == "CRP" ~ "L",
77		#' PARAMCD == "IGA" ~ "H"
78		#' )
79		#' ) %>%
80		#' filter(SAFFL == "Y" & ONTRTFL == "Y" & GRADDR != "")
81		#'
82		#' df <- h_adlb_worsen(
83		#' adlb,
84		#' worst_flag_low = c("WGRLOFL" = "Y"),
85		#' worst_flag_high = c("WGRHIFL" = "Y"),
86		#' direction_var = "GRADDR"
87		#' )
88		#'
89		#' @export
90		h_adlb_worsen <- function(adlb,
91		worst_flag_low = NULL,
92		worst_flag_high = NULL,
93		direction_var) {
94	5x	checkmate::assert_string(direction_var)
95	5x	checkmate::assert_subset(as.character(unique(adlb[[direction_var]])), c("B", "L", "H"))
96	5x	assert_df_with_variables(adlb, list("Col" = direction_var))
97
98	5x	if (any(unique(adlb[[direction_var]]) == "H")) {
99	4x	assert_df_with_variables(adlb, list("High" = names(worst_flag_high)))
100		}
101
102	5x	if (any(unique(adlb[[direction_var]]) == "L")) {
103	4x	assert_df_with_variables(adlb, list("Low" = names(worst_flag_low)))
104		}
105
106	5x	if (any(unique(adlb[[direction_var]]) == "B")) {
107	3x	assert_df_with_variables(
108	3x	adlb,
109	3x	list(
110	3x	"Low" = names(worst_flag_low),
111	3x	"High" = names(worst_flag_high)
112		)
113		)
114		}
115
116		# extract patients with worst post-baseline lab, either low or high or both
117	5x	worst_flag <- c(worst_flag_low, worst_flag_high)
118	5x	col_names <- names(worst_flag)
119	5x	filter_values <- worst_flag
120	5x	temp <- Map(
121	5x	function(x, y) which(adlb[[x]] == y),
122	5x	col_names,
123	5x	filter_values
124		)
125	5x	position_satisfy_filters <- Reduce(union, temp)
126
127		# select variables of interest
128	5x	adlb_f <- adlb[position_satisfy_filters, ]
129
130		# generate subsets for different directionality
131	5x	adlb_f_h <- adlb_f[which(adlb_f[[direction_var]] == "H"), ]
132	5x	adlb_f_l <- adlb_f[which(adlb_f[[direction_var]] == "L"), ]
133	5x	adlb_f_b <- adlb_f[which(adlb_f[[direction_var]] == "B"), ]
134
135		# for labs requiring both high and low, data is duplicated and will be stacked on top of each other
136	5x	adlb_f_b_h <- adlb_f_b
137	5x	adlb_f_b_l <- adlb_f_b
138
139		# extract data with worst lab
140	5x	if (!is.null(worst_flag_high) && !is.null(worst_flag_low)) {
141		# change H to High, L to Low
142	3x	adlb_f_h[[direction_var]] <- rep("High", nrow(adlb_f_h))
143	3x	adlb_f_l[[direction_var]] <- rep("Low", nrow(adlb_f_l))
144
145		# change, B to High and Low
146	3x	adlb_f_b_h[[direction_var]] <- rep("High", nrow(adlb_f_b_h))
147	3x	adlb_f_b_l[[direction_var]] <- rep("Low", nrow(adlb_f_b_l))
148
149	3x	adlb_out_h <- adlb_f_h[which(adlb_f_h[[names(worst_flag_high)]] == worst_flag_high), ]
150	3x	adlb_out_b_h <- adlb_f_b_h[which(adlb_f_b_h[[names(worst_flag_high)]] == worst_flag_high), ]
151	3x	adlb_out_l <- adlb_f_l[which(adlb_f_l[[names(worst_flag_low)]] == worst_flag_low), ]
152	3x	adlb_out_b_l <- adlb_f_b_l[which(adlb_f_b_l[[names(worst_flag_low)]] == worst_flag_low), ]
153
154	3x	out <- rbind(adlb_out_h, adlb_out_b_h, adlb_out_l, adlb_out_b_l)
155	2x	} else if (!is.null(worst_flag_high)) {
156	1x	adlb_f_h[[direction_var]] <- rep("High", nrow(adlb_f_h))
157	1x	adlb_f_b_h[[direction_var]] <- rep("High", nrow(adlb_f_b_h))
158
159	1x	adlb_out_h <- adlb_f_h[which(adlb_f_h[[names(worst_flag_high)]] == worst_flag_high), ]
160	1x	adlb_out_b_h <- adlb_f_b_h[which(adlb_f_b_h[[names(worst_flag_high)]] == worst_flag_high), ]
161
162	1x	out <- rbind(adlb_out_h, adlb_out_b_h)
163	1x	} else if (!is.null(worst_flag_low)) {
164	1x	adlb_f_l[[direction_var]] <- rep("Low", nrow(adlb_f_l))
165	1x	adlb_f_b_l[[direction_var]] <- rep("Low", nrow(adlb_f_b_l))
166
167	1x	adlb_out_l <- adlb_f_l[which(adlb_f_l[[names(worst_flag_low)]] == worst_flag_low), ]
168	1x	adlb_out_b_l <- adlb_f_b_l[which(adlb_f_b_l[[names(worst_flag_low)]] == worst_flag_low), ]
169
170	1x	out <- rbind(adlb_out_l, adlb_out_b_l)
171		}
172
173		# label
174	5x	formatters::var_labels(out) <- formatters::var_labels(adlb_f, fill = FALSE)
175		# NA
176	5x	out
177		}
178
179		#' Helper function to analyze patients for `s_count_abnormal_lab_worsen_by_baseline()`
180		#'
181		#' @description `r lifecycle::badge("stable")`
182		#'
183		#' Helper function to count the number of patients and the fraction of patients according to
184		#' highest post-baseline lab grade variable `.var`, baseline lab grade variable `baseline_var`,
185		#' and the direction of interest specified in `direction_var`.
186		#'
187		#' @inheritParams argument_convention
188		#' @inheritParams h_adlb_worsen
189		#' @param baseline_var (`string`)\cr name of the baseline lab grade variable.
190		#'
191		#' @return The counts and fraction of patients
192		#' whose worst post-baseline lab grades are worse than their baseline grades, for
193		#' post-baseline worst grades "1", "2", "3", "4" and "Any".
194		#'
195		#' @seealso [abnormal_by_worst_grade_worsen]
196		#'
197		#' @examples
198		#' library(dplyr)
199		#'
200		#' # The direction variable, GRADDR, is based on metadata
201		#' adlb <- tern_ex_adlb %>%
202		#' mutate(
203		#' GRADDR = case_when(
204		#' PARAMCD == "ALT" ~ "B",
205		#' PARAMCD == "CRP" ~ "L",
206		#' PARAMCD == "IGA" ~ "H"
207		#' )
208		#' ) %>%
209		#' filter(SAFFL == "Y" & ONTRTFL == "Y" & GRADDR != "")
210		#'
211		#' df <- h_adlb_worsen(
212		#' adlb,
213		#' worst_flag_low = c("WGRLOFL" = "Y"),
214		#' worst_flag_high = c("WGRHIFL" = "Y"),
215		#' direction_var = "GRADDR"
216		#' )
217		#'
218		#' # `h_worsen_counter`
219		#' h_worsen_counter(
220		#' df %>% filter(PARAMCD == "CRP" & GRADDR == "Low"),
221		#' id = "USUBJID",
222		#' .var = "ATOXGR",
223		#' baseline_var = "BTOXGR",
224		#' direction_var = "GRADDR"
225		#' )
226		#'
227		#' @export
228		h_worsen_counter <- function(df, id, .var, baseline_var, direction_var) {
229	17x	checkmate::assert_string(id)
230	17x	checkmate::assert_string(.var)
231	17x	checkmate::assert_string(baseline_var)
232	17x	checkmate::assert_scalar(unique(df[[direction_var]]))
233	17x	checkmate::assert_subset(unique(df[[direction_var]]), c("High", "Low"))
234	17x	assert_df_with_variables(df, list(val = c(id, .var, baseline_var, direction_var)))
235
236		# remove post-baseline missing
237	17x	df <- df[df[[.var]] != "<Missing>", ]
238
239		# obtain directionality
240	17x	direction <- unique(df[[direction_var]])
241
242	17x	if (direction == "Low") {
243	10x	grade <- -1:-4
244	10x	worst_grade <- -4
245	7x	} else if (direction == "High") {
246	7x	grade <- 1:4
247	7x	worst_grade <- 4
248		}
249
250	17x	if (nrow(df) > 0) {
251	17x	by_grade <- lapply(grade, function(i) {
252		# filter baseline values that is less than i or <Missing>
253	68x	df_temp <- df[df[[baseline_var]] %in% c((i + sign(i) * -1):(-1 * worst_grade), "<Missing>"), ]
254		# num: number of patients with post-baseline worst lab equal to i
255	68x	num <- length(unique(df_temp[df_temp[[.var]] %in% i, id, drop = TRUE]))
256		# denom: number of patients with baseline values less than i or <missing> and post-baseline in the same direction
257	68x	denom <- length(unique(df_temp[[id]]))
258	68x	rm(df_temp)
259	68x	c(num = num, denom = denom)
260		})
261		} else {
262	!	by_grade <- lapply(1, function(i) {
263	!	c(num = 0, denom = 0)
264		})
265		}
266
267	17x	names(by_grade) <- as.character(seq_along(by_grade))
268
269		# baseline grade less 4 or missing
270	17x	df_temp <- df[!df[[baseline_var]] %in% worst_grade, ]
271
272		# denom: number of patients with baseline values less than 4 or <missing> and post-baseline in the same direction
273	17x	denom <- length(unique(df_temp[, id, drop = TRUE]))
274
275		# condition 1: missing baseline and in the direction of abnormality
276	17x	con1 <- which(df_temp[[baseline_var]] == "<Missing>" & df_temp[[.var]] %in% grade)
277	17x	df_temp_nm <- df_temp[which(df_temp[[baseline_var]] != "<Missing>" & df_temp[[.var]] %in% grade), ]
278
279		# condition 2: if post-baseline values are present then post-baseline values must be worse than baseline
280	17x	if (direction == "Low") {
281	10x	con2 <- which(as.numeric(as.character(df_temp_nm[[.var]])) < as.numeric(as.character(df_temp_nm[[baseline_var]])))
282		} else {
283	7x	con2 <- which(as.numeric(as.character(df_temp_nm[[.var]])) > as.numeric(as.character(df_temp_nm[[baseline_var]])))
284		}
285
286		# number of patients satisfy either conditions 1 or 2
287	17x	num <- length(unique(df_temp[union(con1, con2), id, drop = TRUE]))
288
289	17x	list(fraction = c(by_grade, list("Any" = c(num = num, denom = denom))))
290		}
291
292		#' @describeIn abnormal_by_worst_grade_worsen Statistics function for patients whose worst post-baseline
293		#' lab grades are worse than their baseline grades.
294		#'
295		#' @return
296		#' * `s_count_abnormal_lab_worsen_by_baseline()` returns the counts and fraction of patients whose worst
297		#' post-baseline lab grades are worse than their baseline grades, for post-baseline worst grades
298		#' "1", "2", "3", "4" and "Any".
299		#'
300		#' @keywords internal
301		s_count_abnormal_lab_worsen_by_baseline <- function(df, # nolint
302		.var = "ATOXGR",
303		variables = list(
304		id = "USUBJID",
305		baseline_var = "BTOXGR",
306		direction_var = "GRADDR"
307		)) {
308	1x	checkmate::assert_string(.var)
309	1x	checkmate::assert_set_equal(names(variables), c("id", "baseline_var", "direction_var"))
310	1x	checkmate::assert_string(variables$id)
311	1x	checkmate::assert_string(variables$baseline_var)
312	1x	checkmate::assert_string(variables$direction_var)
313	1x	assert_df_with_variables(df, c(aval = .var, variables[1:3]))
314	1x	assert_list_of_variables(variables)
315
316	1x	h_worsen_counter(df, variables$id, .var, variables$baseline_var, variables$direction_var)
317		}
318
319		#' @describeIn abnormal_by_worst_grade_worsen Formatted analysis function which is used as `afun`
320		#' in `count_abnormal_lab_worsen_by_baseline()`.
321		#'
322		#' @return
323		#' * `a_count_abnormal_lab_worsen_by_baseline()` returns the corresponding list with
324		#' formatted [rtables::CellValue()].
325		#'
326		#' @keywords internal
327		a_count_abnormal_lab_worsen_by_baseline <- make_afun( # nolint
328		s_count_abnormal_lab_worsen_by_baseline,
329		.formats = c(fraction = format_fraction),
330		.ungroup_stats = "fraction"
331		)
332
333		#' @describeIn abnormal_by_worst_grade_worsen Layout-creating function which can take statistics function
334		#' arguments and additional format arguments. This function is a wrapper for [rtables::analyze()].
335		#'
336		#' @return
337		#' * `count_abnormal_lab_worsen_by_baseline()` returns a layout object suitable for passing to further layouting
338		#' functions, or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted
339		#' rows containing the statistics from `s_count_abnormal_lab_worsen_by_baseline()` to the table layout.
340		#'
341		#' @examples
342		#' library(dplyr)
343		#'
344		#' # The direction variable, GRADDR, is based on metadata
345		#' adlb <- tern_ex_adlb %>%
346		#' mutate(
347		#' GRADDR = case_when(
348		#' PARAMCD == "ALT" ~ "B",
349		#' PARAMCD == "CRP" ~ "L",
350		#' PARAMCD == "IGA" ~ "H"
351		#' )
352		#' ) %>%
353		#' filter(SAFFL == "Y" & ONTRTFL == "Y" & GRADDR != "")
354		#'
355		#' df <- h_adlb_worsen(
356		#' adlb,
357		#' worst_flag_low = c("WGRLOFL" = "Y"),
358		#' worst_flag_high = c("WGRHIFL" = "Y"),
359		#' direction_var = "GRADDR"
360		#' )
361		#'
362		#' basic_table() %>%
363		#' split_cols_by("ARMCD") %>%
364		#' add_colcounts() %>%
365		#' split_rows_by("PARAMCD") %>%
366		#' split_rows_by("GRADDR") %>%
367		#' count_abnormal_lab_worsen_by_baseline(
368		#' var = "ATOXGR",
369		#' variables = list(
370		#' id = "USUBJID",
371		#' baseline_var = "BTOXGR",
372		#' direction_var = "GRADDR"
373		#' )
374		#' ) %>%
375		#' append_topleft("Direction of Abnormality") %>%
376		#' build_table(df = df, alt_counts_df = tern_ex_adsl)
377		#'
378		#' @export
379		#' @order 2
380		count_abnormal_lab_worsen_by_baseline <- function(lyt, # nolint
381		var,
382		variables = list(
383		id = "USUBJID",
384		baseline_var = "BTOXGR",
385		direction_var = "GRADDR"
386		),
387		na_str = default_na_str(),
388		nested = TRUE,
389		...,
390		table_names = NULL,
391		.stats = NULL,
392		.formats = NULL,
393		.labels = NULL,
394		.indent_mods = NULL) {
395	1x	checkmate::assert_string(var)
396
397	1x	extra_args <- list(variables = variables, ...)
398
399	1x	afun <- make_afun(
400	1x	a_count_abnormal_lab_worsen_by_baseline,
401	1x	.stats = .stats,
402	1x	.formats = .formats,
403	1x	.labels = .labels,
404	1x	.indent_mods = .indent_mods
405		)
406
407	1x	lyt <- analyze(
408	1x	lyt = lyt,
409	1x	vars = var,
410	1x	afun = afun,
411	1x	na_str = na_str,
412	1x	nested = nested,
413	1x	extra_args = extra_args,
414	1x	show_labels = "hidden"
415		)
416
417	1x	lyt
418		}

1		#' Helper function for deriving analysis datasets for select laboratory tables
2		#'
3		#' @description `r lifecycle::badge("stable")`
4		#'
5		#' Helper function that merges ADSL and ADLB datasets so that missing lab test records are inserted in the
6		#' output dataset. Remember that `na_level` must match the needed pre-processing
7		#' done with [df_explicit_na()] to have the desired output.
8		#'
9		#' @param adsl (`data.frame`)\cr ADSL data frame.
10		#' @param adlb (`data.frame`)\cr ADLB data frame.
11		#' @param worst_flag (named `character`)\cr worst post-baseline lab flag variable. See how this is implemented in the
12		#' following examples.
13		#' @param by_visit (`flag`)\cr defaults to `FALSE` to generate worst grade per patient.
14		#' If worst grade per patient per visit is specified for `worst_flag`, then
15		#' `by_visit` should be `TRUE` to generate worst grade patient per visit.
16		#' @param no_fillin_visits (named `character`)\cr visits that are not considered for post-baseline worst toxicity
17		#' grade. Defaults to `c("SCREENING", "BASELINE")`.
18		#'
19		#' @return `df` containing variables shared between `adlb` and `adsl` along with variables `PARAM`, `PARAMCD`,
20		#' `ATOXGR`, and `BTOXGR` relevant for analysis. Optionally, `AVISIT` are `AVISITN` are included when
21		#' `by_visit = TRUE` and `no_fillin_visits = c("SCREENING", "BASELINE")`.
22		#'
23		#' @details In the result data missing records will be created for the following situations:
24		#' * Patients who are present in `adsl` but have no lab data in `adlb` (both baseline and post-baseline).
25		#' * Patients who do not have any post-baseline lab values.
26		#' * Patients without any post-baseline values flagged as the worst.
27		#'
28		#' @examples
29		#' # `h_adsl_adlb_merge_using_worst_flag`
30		#' adlb_out <- h_adsl_adlb_merge_using_worst_flag(
31		#' tern_ex_adsl,
32		#' tern_ex_adlb,
33		#' worst_flag = c("WGRHIFL" = "Y")
34		#' )
35		#'
36		#' # `h_adsl_adlb_merge_using_worst_flag` by visit example
37		#' adlb_out_by_visit <- h_adsl_adlb_merge_using_worst_flag(
38		#' tern_ex_adsl,
39		#' tern_ex_adlb,
40		#' worst_flag = c("WGRLOVFL" = "Y"),
41		#' by_visit = TRUE
42		#' )
43		#'
44		#' @export
45		h_adsl_adlb_merge_using_worst_flag <- function(adsl, # nolint
46		adlb,
47		worst_flag = c("WGRHIFL" = "Y"),
48		by_visit = FALSE,
49		no_fillin_visits = c("SCREENING", "BASELINE")) {
50	5x	col_names <- names(worst_flag)
51	5x	filter_values <- worst_flag
52
53	5x	temp <- Map(
54	5x	function(x, y) which(adlb[[x]] == y),
55	5x	col_names,
56	5x	filter_values
57		)
58
59	5x	position_satisfy_filters <- Reduce(intersect, temp)
60
61	5x	adsl_adlb_common_columns <- intersect(colnames(adsl), colnames(adlb))
62	5x	columns_from_adlb <- c("USUBJID", "PARAM", "PARAMCD", "AVISIT", "AVISITN", "ATOXGR", "BTOXGR")
63
64	5x	adlb_f <- adlb[position_satisfy_filters, ] %>%
65	5x	dplyr::filter(!.data[["AVISIT"]] %in% no_fillin_visits)
66	5x	adlb_f <- adlb_f[, columns_from_adlb]
67
68	5x	avisits_grid <- adlb %>%
69	5x	dplyr::filter(!.data[["AVISIT"]] %in% no_fillin_visits) %>%
70	5x	dplyr::pull(.data[["AVISIT"]]) %>%
71	5x	unique()
72
73	5x	if (by_visit) {
74	1x	adsl_lb <- expand.grid(
75	1x	USUBJID = unique(adsl$USUBJID),
76	1x	AVISIT = avisits_grid,
77	1x	PARAMCD = unique(adlb$PARAMCD)
78		)
79
80	1x	adsl_lb <- adsl_lb %>%
81	1x	dplyr::left_join(unique(adlb[c("AVISIT", "AVISITN")]), by = "AVISIT") %>%
82	1x	dplyr::left_join(unique(adlb[c("PARAM", "PARAMCD")]), by = "PARAMCD")
83
84	1x	adsl1 <- adsl[, adsl_adlb_common_columns]
85	1x	adsl_lb <- adsl1 %>% merge(adsl_lb, by = "USUBJID")
86
87	1x	by_variables_from_adlb <- c("USUBJID", "AVISIT", "AVISITN", "PARAMCD", "PARAM")
88
89	1x	adlb_btoxgr <- adlb %>%
90	1x	dplyr::select(c("USUBJID", "PARAMCD", "BTOXGR")) %>%
91	1x	unique() %>%
92	1x	dplyr::rename("BTOXGR_MAP" = "BTOXGR")
93
94	1x	adlb_out <- merge(
95	1x	adlb_f,
96	1x	adsl_lb,
97	1x	by = by_variables_from_adlb,
98	1x	all = TRUE,
99	1x	sort = FALSE
100		)
101	1x	adlb_out <- adlb_out %>%
102	1x	dplyr::left_join(adlb_btoxgr, by = c("USUBJID", "PARAMCD")) %>%
103	1x	dplyr::mutate(BTOXGR = .data$BTOXGR_MAP) %>%
104	1x	dplyr::select(-"BTOXGR_MAP")
105
106	1x	adlb_var_labels <- c(
107	1x	formatters::var_labels(adlb[by_variables_from_adlb]),
108	1x	formatters::var_labels(adlb[columns_from_adlb[!columns_from_adlb %in% by_variables_from_adlb]]),
109	1x	formatters::var_labels(adsl[adsl_adlb_common_columns[adsl_adlb_common_columns != "USUBJID"]])
110		)
111		} else {
112	4x	adsl_lb <- expand.grid(
113	4x	USUBJID = unique(adsl$USUBJID),
114	4x	PARAMCD = unique(adlb$PARAMCD)
115		)
116
117	4x	adsl_lb <- adsl_lb %>% dplyr::left_join(unique(adlb[c("PARAM", "PARAMCD")]), by = "PARAMCD")
118
119	4x	adsl1 <- adsl[, adsl_adlb_common_columns]
120	4x	adsl_lb <- adsl1 %>% merge(adsl_lb, by = "USUBJID")
121
122	4x	by_variables_from_adlb <- c("USUBJID", "PARAMCD", "PARAM")
123
124	4x	adlb_out <- merge(
125	4x	adlb_f,
126	4x	adsl_lb,
127	4x	by = by_variables_from_adlb,
128	4x	all = TRUE,
129	4x	sort = FALSE
130		)
131
132	4x	adlb_var_labels <- c(
133	4x	formatters::var_labels(adlb[by_variables_from_adlb]),
134	4x	formatters::var_labels(adlb[columns_from_adlb[!columns_from_adlb %in% by_variables_from_adlb]]),
135	4x	formatters::var_labels(adsl[adsl_adlb_common_columns[adsl_adlb_common_columns != "USUBJID"]])
136		)
137		}
138
139	5x	adlb_out$ATOXGR <- as.factor(adlb_out$ATOXGR)
140	5x	adlb_out$BTOXGR <- as.factor(adlb_out$BTOXGR)
141
142	5x	formatters::var_labels(adlb_out) <- adlb_var_labels
143
144	5x	adlb_out
145		}

1		#' Count patients with marked laboratory abnormalities
2		#'
3		#' @description `r lifecycle::badge("stable")`
4		#'
5		#' The analyze function [count_abnormal_by_marked()] creates a layout element to count patients with marked laboratory
6		#' abnormalities for each direction of abnormality, categorized by parameter value.
7		#'
8		#' This function analyzes primary analysis variable `var` which indicates whether a single, replicated,
9		#' or last marked laboratory abnormality was observed. Levels of `var` to include for each marked lab
10		#' abnormality (`single` and `last_replicated`) can be supplied via the `category` parameter. Additional
11		#' analysis variables that can be supplied as a list via the `variables` parameter are `id` (defaults
12		#' to `USUBJID`), a variable to indicate unique subject identifiers, `param` (defaults to `PARAM`), a
13		#' variable to indicate parameter values, and `direction` (defaults to `abn_dir`), a variable to indicate
14		#' abnormality directions.
15		#'
16		#' For each combination of `param` and `direction` levels, marked lab abnormality counts are calculated
17		#' as follows:
18		#' * `Single, not last` & `Last or replicated`: The number of patients with `Single, not last`
19		#' and `Last or replicated` values, respectively.
20		#' * `Any`: The number of patients with either single or replicated marked abnormalities.
21		#'
22		#' Fractions are calculated by dividing the above counts by the number of patients with at least one
23		#' valid measurement recorded during the analysis.
24		#'
25		#' Prior to using this function in your table layout you must use [rtables::split_rows_by()] to create two
26		#' row splits, one on variable `param` and one on variable `direction`.
27		#'
28		#' @inheritParams argument_convention
29		#' @param category (`list`)\cr a list with different marked category names for single
30		#' and last or replicated.
31		#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("abnormal_by_marked")`
32		#' to see available statistics for this function.
33		#'
34		#' @note `Single, not last` and `Last or replicated` levels are mutually exclusive. If a patient has
35		#' abnormalities that meet both the `Single, not last` and `Last or replicated` criteria, then the
36		#' patient will be counted only under the `Last or replicated` category.
37		#'
38		#' @name abnormal_by_marked
39		#' @order 1
40		NULL
41
42		#' @describeIn abnormal_by_marked Statistics function for patients with marked lab abnormalities.
43		#'
44		#' @return
45		#' * `s_count_abnormal_by_marked()` returns statistic `count_fraction` with `Single, not last`,
46		#' `Last or replicated`, and `Any` results.
47		#'
48		#' @keywords internal
49		s_count_abnormal_by_marked <- function(df,
50		.var = "AVALCAT1",
51		.spl_context,
52		category = list(single = "SINGLE", last_replicated = c("LAST", "REPLICATED")),
53		variables = list(id = "USUBJID", param = "PARAM", direction = "abn_dir")) {
54	3x	checkmate::assert_string(.var)
55	3x	checkmate::assert_list(variables)
56	3x	checkmate::assert_list(category)
57	3x	checkmate::assert_subset(names(category), c("single", "last_replicated"))
58	3x	checkmate::assert_subset(names(variables), c("id", "param", "direction"))
59	3x	checkmate::assert_vector(unique(df[[variables$direction]]), max.len = 1)
60
61	2x	assert_df_with_variables(df, c(aval = .var, variables))
62	2x	checkmate::assert_multi_class(df[[.var]], classes = c("factor", "character"))
63	2x	checkmate::assert_multi_class(df[[variables$id]], classes = c("factor", "character"))
64
65
66	2x	first_row <- .spl_context[.spl_context$split == variables[["param"]], ]
67		# Patients in the denominator have at least one post-baseline visit.
68	2x	subj <- first_row$full_parent_df[[1]][[variables[["id"]]]]
69	2x	subj_cur_col <- subj[first_row$cur_col_subset[[1]]]
70		# Some subjects may have a record for high and low directions but
71		# should be counted only once.
72	2x	denom <- length(unique(subj_cur_col))
73
74	2x	if (denom != 0) {
75	2x	subjects_last_replicated <- unique(
76	2x	df[df[[.var]] %in% category[["last_replicated"]], variables$id, drop = TRUE]
77		)
78	2x	subjects_single <- unique(
79	2x	df[df[[.var]] %in% category[["single"]], variables$id, drop = TRUE]
80		)
81		# Subjects who have both single and last/replicated abnormalities are counted in only the last/replicated group.
82	2x	subjects_single <- setdiff(subjects_single, subjects_last_replicated)
83	2x	n_single <- length(subjects_single)
84	2x	n_last_replicated <- length(subjects_last_replicated)
85	2x	n_any <- n_single + n_last_replicated
86	2x	result <- list(count_fraction = list(
87	2x	"Single, not last" = c(n_single, n_single / denom),
88	2x	"Last or replicated" = c(n_last_replicated, n_last_replicated / denom),
89	2x	"Any Abnormality" = c(n_any, n_any / denom)
90		))
91		} else {
92	!	result <- list(count_fraction = list(
93	!	"Single, not last" = c(0, 0),
94	!	"Last or replicated" = c(0, 0),
95	!	"Any Abnormality" = c(0, 0)
96		))
97		}
98
99	2x	result
100		}
101
102		#' @describeIn abnormal_by_marked Formatted analysis function which is used as `afun`
103		#' in `count_abnormal_by_marked()`.
104		#'
105		#' @return
106		#' * `a_count_abnormal_by_marked()` returns the corresponding list with formatted [rtables::CellValue()].
107		#'
108		#' @keywords internal
109		a_count_abnormal_by_marked <- make_afun(
110		s_count_abnormal_by_marked,
111		.formats = c(count_fraction = format_count_fraction)
112		)
113
114		#' @describeIn abnormal_by_marked Layout-creating function which can take statistics function arguments
115		#' and additional format arguments. This function is a wrapper for [rtables::analyze()].
116		#'
117		#' @return
118		#' * `count_abnormal_by_marked()` returns a layout object suitable for passing to further layouting functions,
119		#' or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
120		#' the statistics from `s_count_abnormal_by_marked()` to the table layout.
121		#'
122		#' @examples
123		#' library(dplyr)
124		#'
125		#' df <- data.frame(
126		#' USUBJID = as.character(c(rep(1, 5), rep(2, 5), rep(1, 5), rep(2, 5))),
127		#' ARMCD = factor(c(rep("ARM A", 5), rep("ARM B", 5), rep("ARM A", 5), rep("ARM B", 5))),
128		#' ANRIND = factor(c(
129		#' "NORMAL", "HIGH", "HIGH", "HIGH HIGH", "HIGH",
130		#' "HIGH", "HIGH", "HIGH HIGH", "NORMAL", "HIGH HIGH", "NORMAL", "LOW", "LOW", "LOW LOW", "LOW",
131		#' "LOW", "LOW", "LOW LOW", "NORMAL", "LOW LOW"
132		#' )),
133		#' ONTRTFL = rep(c("", "Y", "Y", "Y", "Y", "Y", "Y", "Y", "Y", "Y"), 2),
134		#' PARAMCD = factor(c(rep("CRP", 10), rep("ALT", 10))),
135		#' AVALCAT1 = factor(rep(c("", "", "", "SINGLE", "REPLICATED", "", "", "LAST", "", "SINGLE"), 2)),
136		#' stringsAsFactors = FALSE
137		#' )
138		#'
139		#' df <- df %>%
140		#' mutate(abn_dir = factor(
141		#' case_when(
142		#' ANRIND == "LOW LOW" ~ "Low",
143		#' ANRIND == "HIGH HIGH" ~ "High",
144		#' TRUE ~ ""
145		#' ),
146		#' levels = c("Low", "High")
147		#' ))
148		#'
149		#' # Select only post-baseline records.
150		#' df <- df %>% filter(ONTRTFL == "Y")
151		#' df_crp <- df %>%
152		#' filter(PARAMCD == "CRP") %>%
153		#' droplevels()
154		#' full_parent_df <- list(df_crp, "not_needed")
155		#' cur_col_subset <- list(rep(TRUE, nrow(df_crp)), "not_needed")
156		#' spl_context <- data.frame(
157		#' split = c("PARAMCD", "GRADE_DIR"),
158		#' full_parent_df = I(full_parent_df),
159		#' cur_col_subset = I(cur_col_subset)
160		#' )
161		#'
162		#' map <- unique(
163		#' df[df$abn_dir %in% c("Low", "High") & df$AVALCAT1 != "", c("PARAMCD", "abn_dir")]
164		#' ) %>%
165		#' lapply(as.character) %>%
166		#' as.data.frame() %>%
167		#' arrange(PARAMCD, abn_dir)
168		#'
169		#' basic_table() %>%
170		#' split_cols_by("ARMCD") %>%
171		#' split_rows_by("PARAMCD") %>%
172		#' summarize_num_patients(
173		#' var = "USUBJID",
174		#' .stats = "unique_count"
175		#' ) %>%
176		#' split_rows_by(
177		#' "abn_dir",
178		#' split_fun = trim_levels_to_map(map)
179		#' ) %>%
180		#' count_abnormal_by_marked(
181		#' var = "AVALCAT1",
182		#' variables = list(
183		#' id = "USUBJID",
184		#' param = "PARAMCD",
185		#' direction = "abn_dir"
186		#' )
187		#' ) %>%
188		#' build_table(df = df)
189		#'
190		#' basic_table() %>%
191		#' split_cols_by("ARMCD") %>%
192		#' split_rows_by("PARAMCD") %>%
193		#' summarize_num_patients(
194		#' var = "USUBJID",
195		#' .stats = "unique_count"
196		#' ) %>%
197		#' split_rows_by(
198		#' "abn_dir",
199		#' split_fun = trim_levels_in_group("abn_dir")
200		#' ) %>%
201		#' count_abnormal_by_marked(
202		#' var = "AVALCAT1",
203		#' variables = list(
204		#' id = "USUBJID",
205		#' param = "PARAMCD",
206		#' direction = "abn_dir"
207		#' )
208		#' ) %>%
209		#' build_table(df = df)
210		#'
211		#' @export
212		#' @order 2
213		count_abnormal_by_marked <- function(lyt,
214		var,
215		category = list(single = "SINGLE", last_replicated = c("LAST", "REPLICATED")),
216		variables = list(id = "USUBJID", param = "PARAM", direction = "abn_dir"),
217		na_str = default_na_str(),
218		nested = TRUE,
219		...,
220		.stats = NULL,
221		.formats = NULL,
222		.labels = NULL,
223		.indent_mods = NULL) {
224	1x	checkmate::assert_string(var)
225
226	1x	extra_args <- list(category = category, variables = variables, ...)
227
228	1x	afun <- make_afun(
229	1x	a_count_abnormal_by_marked,
230	1x	.stats = .stats,
231	1x	.formats = .formats,
232	1x	.labels = .labels,
233	1x	.indent_mods = .indent_mods,
234	1x	.ungroup_stats = "count_fraction"
235		)
236
237	1x	lyt <- analyze(
238	1x	lyt = lyt,
239	1x	vars = var,
240	1x	afun = afun,
241	1x	na_str = na_str,
242	1x	nested = nested,
243	1x	show_labels = "hidden",
244	1x	extra_args = extra_args
245		)
246	1x	lyt
247		}

1		#' Formatting functions
2		#'
3		#' @description `r lifecycle::badge("stable")`
4		#'
5		#' See below for the list of formatting functions created in `tern` to work with `rtables`.
6		#'
7		#' Other available formats can be listed via [`formatters::list_valid_format_labels()`]. Additional
8		#' custom formats can be created via the [`formatters::sprintf_format()`] function.
9		#'
10		#' @family formatting functions
11		#' @name formatting_functions
12		NULL
13
14		#' Format fraction and percentage
15		#'
16		#' @description `r lifecycle::badge("stable")`
17		#'
18		#' Formats a fraction together with ratio in percent.
19		#'
20		#' @param x (named `integer`)\cr vector with elements `num` and `denom`.
21		#' @param ... not used. Required for `rtables` interface.
22		#'
23		#' @return A string in the format `num / denom (ratio %)`. If `num` is 0, the format is `num / denom`.
24		#'
25		#' @examples
26		#' format_fraction(x = c(num = 2L, denom = 3L))
27		#' format_fraction(x = c(num = 0L, denom = 3L))
28		#'
29		#' @family formatting functions
30		#' @export
31		format_fraction <- function(x, ...) {
32	4x	attr(x, "label") <- NULL
33
34	4x	checkmate::assert_vector(x)
35	4x	checkmate::assert_count(x["num"])
36	2x	checkmate::assert_count(x["denom"])
37
38	2x	result <- if (x["num"] == 0) {
39	1x	paste0(x["num"], "/", x["denom"])
40		} else {
41	1x	paste0(
42	1x	x["num"], "/", x["denom"],
43	1x	" (", round(x["num"] / x["denom"] * 100, 1), "%)"
44		)
45		}
46
47	2x	return(result)
48		}
49
50		#' Format fraction and percentage with fixed single decimal place
51		#'
52		#' @description `r lifecycle::badge("stable")`
53		#'
54		#' Formats a fraction together with ratio in percent with fixed single decimal place.
55		#' Includes trailing zero in case of whole number percentages to always keep one decimal place.
56		#'
57		#' @inheritParams format_fraction
58		#'
59		#' @return A string in the format `num / denom (ratio %)`. If `num` is 0, the format is `num / denom`.
60		#'
61		#' @examples
62		#' format_fraction_fixed_dp(x = c(num = 1L, denom = 2L))
63		#' format_fraction_fixed_dp(x = c(num = 1L, denom = 4L))
64		#' format_fraction_fixed_dp(x = c(num = 0L, denom = 3L))
65		#'
66		#' @family formatting functions
67		#' @export
68		format_fraction_fixed_dp <- function(x, ...) {
69	3x	attr(x, "label") <- NULL
70	3x	checkmate::assert_vector(x)
71	3x	checkmate::assert_count(x["num"])
72	3x	checkmate::assert_count(x["denom"])
73
74	3x	result <- if (x["num"] == 0) {
75	1x	paste0(x["num"], "/", x["denom"])
76		} else {
77	2x	paste0(
78	2x	x["num"], "/", x["denom"],
79	2x	" (", sprintf("%.1f", round(x["num"] / x["denom"] * 100, 1)), "%)"
80		)
81		}
82	3x	return(result)
83		}
84
85		#' Format count and fraction
86		#'
87		#' @description `r lifecycle::badge("stable")`
88		#'
89		#' Formats a count together with fraction with special consideration when count is `0`.
90		#'
91		#' @param x (`numeric(2)`)\cr vector of length 2 with count and fraction, respectively.
92		#' @param ... not used. Required for `rtables` interface.
93		#'
94		#' @return A string in the format `count (fraction %)`. If `count` is 0, the format is `0`.
95		#'
96		#' @examples
97		#' format_count_fraction(x = c(2, 0.6667))
98		#' format_count_fraction(x = c(0, 0))
99		#'
100		#' @family formatting functions
101		#' @export
102		format_count_fraction <- function(x, ...) {
103	3x	attr(x, "label") <- NULL
104
105	3x	if (any(is.na(x))) {
106	1x	return("NA")
107		}
108
109	2x	checkmate::assert_vector(x)
110	2x	checkmate::assert_integerish(x[1])
111	2x	assert_proportion_value(x[2], include_boundaries = TRUE)
112
113	2x	result <- if (x[1] == 0) {
114	1x	"0"
115		} else {
116	1x	paste0(x[1], " (", round(x[2] * 100, 1), "%)")
117		}
118
119	2x	return(result)
120		}
121
122		#' Format count and percentage with fixed single decimal place
123		#'
124		#' @description `r lifecycle::badge("experimental")`
125		#'
126		#' Formats a count together with fraction with special consideration when count is `0`.
127		#'
128		#' @inheritParams format_count_fraction
129		#'
130		#' @return A string in the format `count (fraction %)`. If `count` is 0, the format is `0`.
131		#'
132		#' @examples
133		#' format_count_fraction_fixed_dp(x = c(2, 0.6667))
134		#' format_count_fraction_fixed_dp(x = c(2, 0.5))
135		#' format_count_fraction_fixed_dp(x = c(0, 0))
136		#'
137		#' @family formatting functions
138		#' @export
139		format_count_fraction_fixed_dp <- function(x, ...) {
140	503x	attr(x, "label") <- NULL
141
142	503x	if (any(is.na(x))) {
143	!	return("NA")
144		}
145
146	503x	checkmate::assert_vector(x)
147	503x	checkmate::assert_integerish(x[1])
148	503x	assert_proportion_value(x[2], include_boundaries = TRUE)
149
150	503x	result <- if (x[1] == 0) {
151	1x	"0"
152	503x	} else if (.is_equal_float(x[2], 1)) {
153	500x	sprintf("%d (100%%)", x[1])
154		} else {
155	2x	sprintf("%d (%.1f%%)", x[1], x[2] * 100)
156		}
157
158	503x	return(result)
159		}
160
161		#' Format count and fraction with special case for count < 10
162		#'
163		#' @description `r lifecycle::badge("stable")`
164		#'
165		#' Formats a count together with fraction with special consideration when count is less than 10.
166		#'
167		#' @inheritParams format_count_fraction
168		#'
169		#' @return A string in the format `count (fraction %)`. If `count` is less than 10, only `count` is printed.
170		#'
171		#' @examples
172		#' format_count_fraction_lt10(x = c(275, 0.9673))
173		#' format_count_fraction_lt10(x = c(2, 0.6667))
174		#' format_count_fraction_lt10(x = c(9, 1))
175		#'
176		#' @family formatting functions
177		#' @export
178		format_count_fraction_lt10 <- function(x, ...) {
179	7x	attr(x, "label") <- NULL
180
181	7x	if (any(is.na(x))) {
182	1x	return("NA")
183		}
184
185	6x	checkmate::assert_vector(x)
186	6x	checkmate::assert_integerish(x[1])
187	6x	assert_proportion_value(x[2], include_boundaries = TRUE)
188
189	6x	result <- if (x[1] < 10) {
190	3x	paste0(x[1])
191		} else {
192	3x	paste0(x[1], " (", round(x[2] * 100, 1), "%)")
193		}
194
195	6x	return(result)
196		}
197
198		#' Format XX as a formatting function
199		#'
200		#' Translate a string where x and dots are interpreted as number place
201		#' holders, and others as formatting elements.
202		#'
203		#' @param str (`string`)\cr template.
204		#'
205		#' @return An `rtables` formatting function.
206		#'
207		#' @examples
208		#' test <- list(c(1.658, 0.5761), c(1e1, 785.6))
209		#'
210		#' z <- format_xx("xx (xx.x)")
211		#' sapply(test, z)
212		#'
213		#' z <- format_xx("xx.x - xx.x")
214		#' sapply(test, z)
215		#'
216		#' z <- format_xx("xx.x, incl. xx.x% NE")
217		#' sapply(test, z)
218		#'
219		#' @family formatting functions
220		#' @export
221		format_xx <- function(str) {
222		# Find position in the string.
223	1x	positions <- gregexpr(pattern = "x+\\.x+\|x+", text = str, perl = TRUE)
224	1x	x_positions <- regmatches(x = str, m = positions)[[1]]
225
226		# Roundings depends on the number of x behind [.].
227	1x	roundings <- lapply(
228	1x	X = x_positions,
229	1x	function(x) {
230	2x	y <- strsplit(split = "\\.", x = x)[[1]]
231	2x	rounding <- function(x) {
232	4x	round(x, digits = ifelse(length(y) > 1, nchar(y[2]), 0))
233		}
234	2x	return(rounding)
235		}
236		)
237
238	1x	rtable_format <- function(x, output) {
239	2x	values <- Map(y = x, fun = roundings, function(y, fun) fun(y))
240	2x	regmatches(x = str, m = positions)[[1]] <- values
241	2x	return(str)
242		}
243
244	1x	return(rtable_format)
245		}
246
247		#' Format numeric values by significant figures
248		#'
249		#' Format numeric values to print with a specified number of significant figures.
250		#'
251		#' @param sigfig (`integer(1)`)\cr number of significant figures to display.
252		#' @param format (`string`)\cr the format label (string) to apply when printing the value. Decimal
253		#' places in string are ignored in favor of formatting by significant figures. Formats options are:
254		#' `"xx"`, `"xx / xx"`, `"(xx, xx)"`, `"xx - xx"`, and `"xx (xx)"`.
255		#' @param num_fmt (`string`)\cr numeric format modifiers to apply to the value. Defaults to `"fg"` for
256		#' standard significant figures formatting - fixed (non-scientific notation) format (`"f"`)
257		#' and `sigfig` equal to number of significant figures instead of decimal places (`"g"`). See the
258		#' [formatC()] `format` argument for more options.
259		#'
260		#' @return An `rtables` formatting function.
261		#'
262		#' @examples
263		#' fmt_3sf <- format_sigfig(3)
264		#' fmt_3sf(1.658)
265		#' fmt_3sf(1e1)
266		#'
267		#' fmt_5sf <- format_sigfig(5)
268		#' fmt_5sf(0.57)
269		#' fmt_5sf(0.000025645)
270		#'
271		#' @family formatting functions
272		#' @export
273		format_sigfig <- function(sigfig, format = "xx", num_fmt = "fg") {
274	3x	checkmate::assert_integerish(sigfig)
275	3x	format <- gsub("xx\\.\|xx\\.x+", "xx", format)
276	3x	checkmate::assert_choice(format, c("xx", "xx / xx", "(xx, xx)", "xx - xx", "xx (xx)"))
277	3x	function(x, ...) {
278	!	if (!is.numeric(x)) stop("`format_sigfig` cannot be used for non-numeric values. Please choose another format.")
279	12x	num <- formatC(signif(x, digits = sigfig), digits = sigfig, format = num_fmt, flag = "#")
280	12x	num <- gsub("\\.$", "", num) # remove trailing "."
281
282	12x	format_value(num, format)
283		}
284		}
285
286		#' Format fraction with lower threshold
287		#'
288		#' @description `r lifecycle::badge("stable")`
289		#'
290		#' Formats a fraction when the second element of the input `x` is the fraction. It applies
291		#' a lower threshold, below which it is just stated that the fraction is smaller than that.
292		#'
293		#' @param threshold (`proportion`)\cr lower threshold.
294		#'
295		#' @return An `rtables` formatting function that takes numeric input `x` where the second
296		#' element is the fraction that is formatted. If the fraction is above or equal to the threshold,
297		#' then it is displayed in percentage. If it is positive but below the threshold, it returns,
298		#' e.g. "<1" if the threshold is `0.01`. If it is zero, then just "0" is returned.
299		#'
300		#' @examples
301		#' format_fun <- format_fraction_threshold(0.05)
302		#' format_fun(x = c(20, 0.1))
303		#' format_fun(x = c(2, 0.01))
304		#' format_fun(x = c(0, 0))
305		#'
306		#' @family formatting functions
307		#' @export
308		format_fraction_threshold <- function(threshold) {
309	1x	assert_proportion_value(threshold)
310	1x	string_below_threshold <- paste0("<", round(threshold * 100))
311	1x	function(x, ...) {
312	3x	assert_proportion_value(x[2], include_boundaries = TRUE)
313	3x	ifelse(
314	3x	x[2] > 0.01,
315	3x	round(x[2] * 100),
316	3x	ifelse(
317	3x	x[2] == 0,
318	3x	"0",
319	3x	string_below_threshold
320		)
321		)
322		}
323		}
324
325		#' Format extreme values
326		#'
327		#' @description `r lifecycle::badge("stable")`
328		#'
329		#' `rtables` formatting functions that handle extreme values.
330		#'
331		#' @param digits (`integer(1)`)\cr number of decimal places to display.
332		#'
333		#' @details For each input, apply a format to the specified number of `digits`. If the value is
334		#' below a threshold, it returns "<0.01" e.g. if the number of `digits` is 2. If the value is
335		#' above a threshold, it returns ">999.99" e.g. if the number of `digits` is 2.
336		#' If it is zero, then returns "0.00".
337		#'
338		#' @family formatting functions
339		#' @name extreme_format
340		NULL
341
342		#' @describeIn extreme_format Internal helper function to calculate the threshold and create formatted strings
343		#' used in Formatting Functions. Returns a list with elements `threshold` and `format_string`.
344		#'
345		#' @return
346		#' * `h_get_format_threshold()` returns a `list` of 2 elements: `threshold`, with `low` and `high` thresholds,
347		#' and `format_string`, with thresholds formatted as strings.
348		#'
349		#' @examples
350		#' h_get_format_threshold(2L)
351		#'
352		#' @export
353		h_get_format_threshold <- function(digits = 2L) {
354	2113x	checkmate::assert_integerish(digits)
355
356	2113x	low_threshold <- 1 / (10 ^ digits) # styler: off
357	2113x	high_threshold <- 1000 - (1 / (10 ^ digits)) # styler: off
358
359	2113x	string_below_threshold <- paste0("<", low_threshold)
360	2113x	string_above_threshold <- paste0(">", high_threshold)
361
362	2113x	list(
363	2113x	"threshold" = c(low = low_threshold, high = high_threshold),
364	2113x	"format_string" = c(low = string_below_threshold, high = string_above_threshold)
365		)
366		}
367
368		#' @describeIn extreme_format Internal helper function to apply a threshold format to a value.
369		#' Creates a formatted string to be used in Formatting Functions.
370		#'
371		#' @param x (`numeric(1)`)\cr value to format.
372		#'
373		#' @return
374		#' * `h_format_threshold()` returns the given value, or if the value is not within the digit threshold the relation
375		#' of the given value to the digit threshold, as a formatted string.
376		#'
377		#' @examples
378		#' h_format_threshold(0.001)
379		#' h_format_threshold(1000)
380		#'
381		#' @export
382		h_format_threshold <- function(x, digits = 2L) {
383	2115x	if (is.na(x)) {
384	4x	return(x)
385		}
386
387	2111x	checkmate::assert_numeric(x, lower = 0)
388
389	2111x	l_fmt <- h_get_format_threshold(digits)
390
391	2111x	result <- if (x < l_fmt$threshold["low"] && 0 < x) {
392	44x	l_fmt$format_string["low"]
393	2111x	} else if (x > l_fmt$threshold["high"]) {
394	99x	l_fmt$format_string["high"]
395		} else {
396	1968x	sprintf(fmt = paste0("%.", digits, "f"), x)
397		}
398
399	2111x	unname(result)
400		}
401
402		#' Format a single extreme value
403		#'
404		#' @description `r lifecycle::badge("stable")`
405		#'
406		#' Create a formatting function for a single extreme value.
407		#'
408		#' @inheritParams extreme_format
409		#'
410		#' @return An `rtables` formatting function that uses threshold `digits` to return a formatted extreme value.
411		#'
412		#' @examples
413		#' format_fun <- format_extreme_values(2L)
414		#' format_fun(x = 0.127)
415		#' format_fun(x = Inf)
416		#' format_fun(x = 0)
417		#' format_fun(x = 0.009)
418		#'
419		#' @family formatting functions
420		#' @export
421		format_extreme_values <- function(digits = 2L) {
422	63x	function(x, ...) {
423	657x	checkmate::assert_scalar(x, na.ok = TRUE)
424
425	657x	h_format_threshold(x = x, digits = digits)
426		}
427		}
428
429		#' Format extreme values part of a confidence interval
430		#'
431		#' @description `r lifecycle::badge("stable")`
432		#'
433		#' Formatting Function for extreme values part of a confidence interval. Values
434		#' are formatted as e.g. "(xx.xx, xx.xx)" if the number of `digits` is 2.
435		#'
436		#' @inheritParams extreme_format
437		#'
438		#' @return An `rtables` formatting function that uses threshold `digits` to return a formatted extreme
439		#' values confidence interval.
440		#'
441		#' @examples
442		#' format_fun <- format_extreme_values_ci(2L)
443		#' format_fun(x = c(0.127, Inf))
444		#' format_fun(x = c(0, 0.009))
445		#'
446		#' @family formatting functions
447		#' @export
448		format_extreme_values_ci <- function(digits = 2L) {
449	71x	function(x, ...) {
450	726x	checkmate::assert_vector(x, len = 2)
451	726x	l_result <- h_format_threshold(x = x[1], digits = digits)
452	726x	h_result <- h_format_threshold(x = x[2], digits = digits)
453
454	726x	paste0("(", l_result, ", ", h_result, ")")
455		}
456		}
457
458		#' Format automatically using data significant digits
459		#'
460		#' @description `r lifecycle::badge("stable")`
461		#'
462		#' Formatting function for the majority of default methods used in [analyze_vars()].
463		#' For non-derived values, the significant digits of data is used (e.g. range), while derived
464		#' values have one more digits (measure of location and dispersion like mean, standard deviation).
465		#' This function can be called internally with "auto" like, for example,
466		#' `.formats = c("mean" = "auto")`. See details to see how this works with the inner function.
467		#'
468		#' @param dt_var (`numeric`)\cr variable data the statistics were calculated from. Used only to
469		#' find significant digits. In [analyze_vars] this comes from `.df_row` (see
470		#' [rtables::additional_fun_params]), and it is the row data after the above row splits. No
471		#' column split is considered.
472		#' @param x_stat (`string`)\cr string indicating the current statistical method used.
473		#'
474		#' @return A string that `rtables` prints in a table cell.
475		#'
476		#' @details
477		#' The internal function is needed to work with `rtables` default structure for
478		#' format functions, i.e. `function(x, ...)`, where is x are results from statistical evaluation.
479		#' It can be more than one element (e.g. for `.stats = "mean_sd"`).
480		#'
481		#' @examples
482		#' x_todo <- c(0.001, 0.2, 0.0011000, 3, 4)
483		#' res <- c(mean(x_todo[1:3]), sd(x_todo[1:3]))
484		#'
485		#' # x is the result coming into the formatting function -> res!!
486		#' format_auto(dt_var = x_todo, x_stat = "mean_sd")(x = res)
487		#' format_auto(x_todo, "range")(x = range(x_todo))
488		#' no_sc_x <- c(0.0000001, 1)
489		#' format_auto(no_sc_x, "range")(x = no_sc_x)
490		#'
491		#' @family formatting functions
492		#' @export
493		format_auto <- function(dt_var, x_stat) {
494	10x	function(x = "", ...) {
495	18x	checkmate::assert_numeric(x, min.len = 1)
496	18x	checkmate::assert_numeric(dt_var, min.len = 1)
497		# Defaults - they may be a param in the future
498	18x	der_stats <- c(
499	18x	"mean", "sd", "se", "median", "geom_mean", "quantiles", "iqr",
500	18x	"mean_sd", "mean_se", "mean_se", "mean_ci", "mean_sei", "mean_sdi",
501	18x	"median_ci"
502		)
503	18x	nonder_stats <- c("n", "range", "min", "max")
504
505		# Safenet for miss-modifications
506	18x	stopifnot(length(intersect(der_stats, nonder_stats)) == 0) # nolint
507	18x	checkmate::assert_choice(x_stat, c(der_stats, nonder_stats))
508
509		# Finds the max number of digits in data
510	18x	detect_dig <- vapply(dt_var, count_decimalplaces, FUN.VALUE = numeric(1)) %>%
511	18x	max()
512
513	18x	if (x_stat %in% der_stats) {
514	8x	detect_dig <- detect_dig + 1
515		}
516
517		# Render input
518	18x	str_vals <- formatC(x, digits = detect_dig, format = "f")
519	18x	def_fmt <- get_formats_from_stats(x_stat)[[x_stat]]
520	18x	str_fmt <- str_extract(def_fmt, invert = FALSE)[[1]]
521	18x	if (length(str_fmt) != length(str_vals)) {
522	2x	stop(
523	2x	"Number of inserted values as result (", length(str_vals),
524	2x	") is not the same as there should be in the default tern formats for ",
525	2x	x_stat, " (-> ", def_fmt, " needs ", length(str_fmt), " values). ",
526	2x	"See tern_default_formats to check all of them."
527		)
528		}
529
530		# Squashing them together
531	16x	inv_str_fmt <- str_extract(def_fmt, invert = TRUE)[[1]]
532	16x	stopifnot(length(inv_str_fmt) == length(str_vals) + 1) # nolint
533
534	16x	out <- vector("character", length = length(inv_str_fmt) + length(str_vals))
535	16x	is_even <- seq_along(out) %% 2 == 0
536	16x	out[is_even] <- str_vals
537	16x	out[!is_even] <- inv_str_fmt
538
539	16x	return(paste0(out, collapse = ""))
540		}
541		}
542
543		# Utility function that could be useful in general
544		str_extract <- function(string, pattern = "xx\|xx\\.\|xx\\.x+", invert = FALSE) {
545	34x	regmatches(string, gregexpr(pattern, string), invert = invert)
546		}
547
548		# Helper function
549		count_decimalplaces <- function(dec) {
550	161x	if (is.na(dec)) {
551	6x	return(0)
552	155x	} else if (abs(dec - round(dec)) > .Machine$double.eps^0.5) { # For precision
553	122x	nchar(strsplit(format(dec, scientific = FALSE, trim = FALSE), ".", fixed = TRUE)[[1]][[2]])
554		} else {
555	33x	return(0)
556		}
557		}
558
559		#' Apply automatic formatting
560		#'
561		#' Checks if any of the listed formats in `.formats` are `"auto"`, and replaces `"auto"` with
562		#' the correct implementation of `format_auto` for the given statistics, data, and variable.
563		#'
564		#' @inheritParams argument_convention
565		#' @param x_stats (named `list`)\cr a named list of statistics where each element corresponds
566		#' to an element in `.formats`, with matching names.
567		#'
568		#' @keywords internal
569		apply_auto_formatting <- function(.formats, x_stats, .df_row, .var) {
570	420x	is_auto_fmt <- vapply(.formats, function(ii) is.character(ii) && ii == "auto", logical(1))
571	420x	if (any(is_auto_fmt)) {
572	3x	auto_stats <- x_stats[is_auto_fmt]
573	3x	var_df <- .df_row[[.var]] # xxx this can be extended for the WHOLE data or single facets
574	3x	.formats[is_auto_fmt] <- lapply(names(auto_stats), format_auto, dt_var = var_df)
575		}
576	420x	.formats
577		}

1		#' Univariate formula special term
2		#'
3		#' @description `r lifecycle::badge("stable")`
4		#'
5		#' The special term `univariate` indicate that the model should be fitted individually for
6		#' every variable included in univariate.
7		#'
8		#' @param x (`character`)\cr a vector of variable names separated by commas.
9		#'
10		#' @return When used within a model formula, produces univariate models for each variable provided.
11		#'
12		#' @details
13		#' If provided alongside with pairwise specification, the model
14		#' `y ~ ARM + univariate(SEX, AGE, RACE)` lead to the study and comparison of the models
15		#' + `y ~ ARM`
16		#' + `y ~ ARM + SEX`
17		#' + `y ~ ARM + AGE`
18		#' + `y ~ ARM + RACE`
19		#'
20		#' @export
21		univariate <- function(x) {
22	2x	structure(x, varname = deparse(substitute(x)))
23		}
24
25		# Get the right-hand-term of a formula
26		rht <- function(x) {
27	4x	checkmate::assert_formula(x)
28	4x	y <- as.character(rev(x)[[1]])
29	4x	return(y)
30		}
31
32		#' Hazard ratio estimation in interactions
33		#'
34		#' This function estimates the hazard ratios between arms when an interaction variable is given with
35		#' specific values.
36		#'
37		#' @param variable,given (`character(2)`)\cr names of the two variables in the interaction. We seek the estimation of
38		#' the levels of `variable` given the levels of `given`.
39		#' @param lvl_var,lvl_given (`character`)\cr corresponding levels given by [levels()].
40		#' @param mmat (named `numeric`) a vector filled with `0`s used as a template to obtain the design matrix.
41		#' @param coef (`numeric`)\cr vector of estimated coefficients.
42		#' @param vcov (`matrix`)\cr variance-covariance matrix of underlying model.
43		#' @param conf_level (`proportion`)\cr confidence level of estimate intervals.
44		#'
45		#' @details Given the cox regression investigating the effect of Arm (A, B, C; reference A)
46		#' and Sex (F, M; reference Female). The model is abbreviated: y ~ Arm + Sex + Arm x Sex.
47		#' The cox regression estimates the coefficients along with a variance-covariance matrix for:
48		#'
49		#' - b1 (arm b), b2 (arm c)
50		#' - b3 (sex m)
51		#' - b4 (arm b: sex m), b5 (arm c: sex m)
52		#'
53		#' Given that I want an estimation of the Hazard Ratio for arm C/sex M, the estimation
54		#' will be given in reference to arm A/Sex M by exp(b2 + b3 + b5)/ exp(b3) = exp(b2 + b5),
55		#' therefore the interaction coefficient is given by b2 + b5 while the standard error is obtained
56		#' as $1.96 * sqrt(Var b2 + Var b5 + 2 * covariance (b2,b5))$ for a confidence level of 0.95.
57		#'
58		#' @return A list of matrices (one per level of variable) with rows corresponding to the combinations of
59		#' `variable` and `given`, with columns:
60		#' * `coef_hat`: Estimation of the coefficient.
61		#' * `coef_se`: Standard error of the estimation.
62		#' * `hr`: Hazard ratio.
63		#' * `lcl, ucl`: Lower/upper confidence limit of the hazard ratio.
64		#'
65		#' @seealso [s_cox_multivariate()].
66		#'
67		#' @examples
68		#' library(dplyr)
69		#' library(survival)
70		#'
71		#' ADSL <- tern_ex_adsl %>%
72		#' filter(SEX %in% c("F", "M"))
73		#'
74		#' adtte <- tern_ex_adtte %>% filter(PARAMCD == "PFS")
75		#' adtte$ARMCD <- droplevels(adtte$ARMCD)
76		#' adtte$SEX <- droplevels(adtte$SEX)
77		#'
78		#' mod <- coxph(
79		#' formula = Surv(time = AVAL, event = 1 - CNSR) ~ (SEX + ARMCD)^2,
80		#' data = adtte
81		#' )
82		#'
83		#' mmat <- stats::model.matrix(mod)[1, ]
84		#' mmat[!mmat == 0] <- 0
85		#'
86		#' @keywords internal
87		estimate_coef <- function(variable, given,
88		lvl_var, lvl_given,
89		coef,
90		mmat,
91		vcov,
92		conf_level = 0.95) {
93	8x	var_lvl <- paste0(variable, lvl_var[-1]) # [-1]: reference level
94	8x	giv_lvl <- paste0(given, lvl_given)
95
96	8x	design_mat <- expand.grid(variable = var_lvl, given = giv_lvl)
97	8x	design_mat <- design_mat[order(design_mat$variable, design_mat$given), ]
98	8x	design_mat <- within(
99	8x	data = design_mat,
100	8x	expr = {
101	8x	inter <- paste0(variable, ":", given)
102	8x	rev_inter <- paste0(given, ":", variable)
103		}
104		)
105
106	8x	split_by_variable <- design_mat$variable
107	8x	interaction_names <- paste(design_mat$variable, design_mat$given, sep = "/")
108
109	8x	design_mat <- apply(
110	8x	X = design_mat, MARGIN = 1, FUN = function(x) {
111	27x	mmat[names(mmat) %in% x[-which(names(x) == "given")]] <- 1
112	27x	return(mmat)
113		}
114		)
115	8x	colnames(design_mat) <- interaction_names
116
117	8x	betas <- as.matrix(coef)
118
119	8x	coef_hat <- t(design_mat) %*% betas
120	8x	dimnames(coef_hat)[2] <- "coef"
121
122	8x	coef_se <- apply(design_mat, 2, function(x) {
123	27x	vcov_el <- as.logical(x)
124	27x	y <- vcov[vcov_el, vcov_el]
125	27x	y <- sum(y)
126	27x	y <- sqrt(y)
127	27x	return(y)
128		})
129
130	8x	q_norm <- stats::qnorm((1 + conf_level) / 2)
131	8x	y <- cbind(coef_hat, `se(coef)` = coef_se)
132
133	8x	y <- apply(y, 1, function(x) {
134	27x	x["hr"] <- exp(x["coef"])
135	27x	x["lcl"] <- exp(x["coef"] - q_norm * x["se(coef)"])
136	27x	x["ucl"] <- exp(x["coef"] + q_norm * x["se(coef)"])
137
138	27x	return(x)
139		})
140
141	8x	y <- t(y)
142	8x	y <- by(y, split_by_variable, identity)
143	8x	y <- lapply(y, as.matrix)
144
145	8x	attr(y, "details") <- paste0(
146	8x	"Estimations of ", variable,
147	8x	" hazard ratio given the level of ", given, " compared to ",
148	8x	variable, " level ", lvl_var[1], "."
149		)
150	8x	return(y)
151		}
152
153		#' `tryCatch` around `car::Anova`
154		#'
155		#' Captures warnings when executing [car::Anova].
156		#'
157		#' @inheritParams car::Anova
158		#'
159		#' @return A list with item `aov` for the result of the model and `error_text` for the captured warnings.
160		#'
161		#' @examples
162		#' # `car::Anova` on cox regression model including strata and expected
163		#' # a likelihood ratio test triggers a warning as only Wald method is
164		#' # accepted.
165		#'
166		#' library(survival)
167		#'
168		#' mod <- coxph(
169		#' formula = Surv(time = futime, event = fustat) ~ factor(rx) + strata(ecog.ps),
170		#' data = ovarian
171		#' )
172		#'
173		#' @keywords internal
174		try_car_anova <- function(mod,
175		test.statistic) { # nolint
176	2x	y <- tryCatch(
177	2x	withCallingHandlers(
178	2x	expr = {
179	2x	warn_text <- c()
180	2x	list(
181	2x	aov = car::Anova(
182	2x	mod,
183	2x	test.statistic = test.statistic,
184	2x	type = "III"
185		),
186	2x	warn_text = warn_text
187		)
188		},
189	2x	warning = function(w) {
190		# If a warning is detected it is handled as "w".
191	!	warn_text <<- trimws(paste0("Warning in `try_car_anova`: ", w))
192
193		# A warning is sometimes expected, then, we want to restart
194		# the execution while ignoring the warning.
195	!	invokeRestart("muffleWarning")
196		}
197		),
198	2x	finally = {
199		}
200		)
201
202	2x	return(y)
203		}
204
205		#' Fit a Cox regression model and ANOVA
206		#'
207		#' The functions derives the effect p-values using [car::Anova()] from [survival::coxph()] results.
208		#'
209		#' @inheritParams t_coxreg
210		#'
211		#' @return A list with items `mod` (results of [survival::coxph()]), `msum` (result of `summary`) and
212		#' `aov` (result of [car::Anova()]).
213		#'
214		#' @noRd
215		fit_n_aov <- function(formula,
216		data = data,
217		conf_level = conf_level,
218		pval_method = c("wald", "likelihood"),
219		...) {
220	1x	pval_method <- match.arg(pval_method)
221
222	1x	environment(formula) <- environment()
223	1x	suppressWarnings({
224		# We expect some warnings due to coxph which fails strict programming.
225	1x	mod <- survival::coxph(formula, data = data, ...)
226	1x	msum <- summary(mod, conf.int = conf_level)
227		})
228
229	1x	aov <- try_car_anova(
230	1x	mod,
231	1x	test.statistic = switch(pval_method,
232	1x	"wald" = "Wald",
233	1x	"likelihood" = "LR"
234		)
235		)
236
237	1x	warn_attr <- aov$warn_text
238	!	if (!is.null(aov$warn_text)) message(warn_attr)
239
240	1x	aov <- aov$aov
241	1x	y <- list(mod = mod, msum = msum, aov = aov)
242	1x	attr(y, "message") <- warn_attr
243
244	1x	return(y)
245		}
246
247		# argument_checks
248		check_formula <- function(formula) {
249	1x	if (!(inherits(formula, "formula"))) {
250	1x	stop("Check `formula`. A formula should resemble `Surv(time = AVAL, event = 1 - CNSR) ~ study_arm(ARMCD)`.")
251		}
252
253	!	invisible()
254		}
255
256		check_covariate_formulas <- function(covariates) {
257	1x	if (!all(vapply(X = covariates, FUN = inherits, what = "formula", FUN.VALUE = TRUE)) \|\| is.null(covariates)) {
258	1x	stop("Check `covariates`, it should be a list of right-hand-term formulas, e.g. list(Age = ~AGE).")
259		}
260
261	!	invisible()
262		}
263
264		name_covariate_names <- function(covariates) {
265	1x	miss_names <- names(covariates) == ""
266	1x	no_names <- is.null(names(covariates))
267	!	if (any(miss_names)) names(covariates)[miss_names] <- vapply(covariates[miss_names], FUN = rht, FUN.VALUE = "name")
268	!	if (no_names) names(covariates) <- vapply(covariates, FUN = rht, FUN.VALUE = "name")
269	1x	return(covariates)
270		}
271
272		check_increments <- function(increments, covariates) {
273	1x	if (!is.null(increments)) {
274	1x	covariates <- vapply(covariates, FUN = rht, FUN.VALUE = "name")
275	1x	lapply(
276	1x	X = names(increments), FUN = function(x) {
277	3x	if (!x %in% covariates) {
278	1x	warning(
279	1x	paste(
280	1x	"Check `increments`, the `increment` for ", x,
281	1x	"doesn't match any names in investigated covariate(s)."
282		)
283		)
284		}
285		}
286		)
287		}
288
289	1x	invisible()
290		}
291
292		#' Multivariate Cox model - summarized results
293		#'
294		#' Analyses based on multivariate Cox model are usually not performed for the Controlled Substance Reporting or
295		#' regulatory documents but serve exploratory purposes only (e.g., for publication). In practice, the model usually
296		#' includes only the main effects (without interaction terms). It produces the hazard ratio estimates for each of the
297		#' covariates included in the model.
298		#' The analysis follows the same principles (e.g., stratified vs. unstratified analysis and tie handling) as the
299		#' usual Cox model analysis. Since there is usually no pre-specified hypothesis testing for such analysis,
300		#' the p.values need to be interpreted with caution. (Statistical Analysis of Clinical Trials Data with R,
301		#' `NEST's bookdown`)
302		#'
303		#' @param formula (`formula`)\cr a formula corresponding to the investigated [survival::Surv()] survival model
304		#' including covariates.
305		#' @param data (`data.frame`)\cr a data frame which includes the variable in formula and covariates.
306		#' @param conf_level (`proportion`)\cr the confidence level for the hazard ratio interval estimations. Default is 0.95.
307		#' @param pval_method (`string`)\cr the method used for the estimation of p-values, should be one of
308		#' `"wald"` (default) or `"likelihood"`.
309		#' @param ... optional parameters passed to [survival::coxph()]. Can include `ties`, a character string specifying the
310		#' method for tie handling, one of `exact` (default), `efron`, `breslow`.
311		#'
312		#' @return A `list` with elements `mod`, `msum`, `aov`, and `coef_inter`.
313		#'
314		#' @details The output is limited to single effect terms. Work in ongoing for estimation of interaction terms
315		#' but is out of scope as defined by the Global Data Standards Repository
316		#' (`GDS_Standard_TLG_Specs_Tables_2.doc`).
317		#'
318		#' @seealso [estimate_coef()].
319		#'
320		#' @examples
321		#' library(dplyr)
322		#'
323		#' adtte <- tern_ex_adtte
324		#' adtte_f <- subset(adtte, PARAMCD == "OS") # _f: filtered
325		#' adtte_f <- filter(
326		#' adtte_f,
327		#' PARAMCD == "OS" &
328		#' SEX %in% c("F", "M") &
329		#' RACE %in% c("ASIAN", "BLACK OR AFRICAN AMERICAN", "WHITE")
330		#' )
331		#' adtte_f$SEX <- droplevels(adtte_f$SEX)
332		#' adtte_f$RACE <- droplevels(adtte_f$RACE)
333		#'
334		#' @keywords internal
335		s_cox_multivariate <- function(formula, data,
336		conf_level = 0.95,
337		pval_method = c("wald", "likelihood"),
338		...) {
339	1x	tf <- stats::terms(formula, specials = c("strata"))
340	1x	covariates <- rownames(attr(tf, "factors"))[-c(1, unlist(attr(tf, "specials")))]
341	1x	lapply(
342	1x	X = covariates,
343	1x	FUN = function(x) {
344	3x	if (is.character(data[[x]])) {
345	1x	data[[x]] <<- as.factor(data[[x]])
346		}
347	3x	invisible()
348		}
349		)
350	1x	pval_method <- match.arg(pval_method)
351
352		# Results directly exported from environment(fit_n_aov) to environment(s_function_draft)
353	1x	y <- fit_n_aov(
354	1x	formula = formula,
355	1x	data = data,
356	1x	conf_level = conf_level,
357	1x	pval_method = pval_method,
358		...
359		)
360	1x	mod <- y$mod
361	1x	aov <- y$aov
362	1x	msum <- y$msum
363	1x	list2env(as.list(y), environment())
364
365	1x	all_term_labs <- attr(mod$terms, "term.labels")
366	1x	term_labs <- all_term_labs[which(attr(mod$terms, "order") == 1)]
367	1x	names(term_labs) <- term_labs
368
369	1x	coef_inter <- NULL
370	1x	if (any(attr(mod$terms, "order") > 1)) {
371	1x	for_inter <- all_term_labs[attr(mod$terms, "order") > 1]
372	1x	names(for_inter) <- for_inter
373	1x	mmat <- stats::model.matrix(mod)[1, ]
374	1x	mmat[!mmat == 0] <- 0
375	1x	mcoef <- stats::coef(mod)
376	1x	mvcov <- stats::vcov(mod)
377
378	1x	estimate_coef_local <- function(variable, given) {
379	6x	estimate_coef(
380	6x	variable, given,
381	6x	coef = mcoef, mmat = mmat, vcov = mvcov, conf_level = conf_level,
382	6x	lvl_var = levels(data[[variable]]), lvl_given = levels(data[[given]])
383		)
384		}
385
386	1x	coef_inter <- lapply(
387	1x	for_inter, function(x) {
388	3x	y <- attr(mod$terms, "factors")[, x]
389	3x	y <- names(y[y > 0])
390	3x	Map(estimate_coef_local, variable = y, given = rev(y))
391		}
392		)
393		}
394
395	1x	list(mod = mod, msum = msum, aov = aov, coef_inter = coef_inter)
396		}

1		#' Proportion estimation
2		#'
3		#' @description `r lifecycle::badge("stable")`
4		#'
5		#' The analyze function [estimate_proportion()] creates a layout element to estimate the proportion of responders
6		#' within a studied population. The primary analysis variable, `vars`, indicates whether a response has occurred for
7		#' each record. See the `method` parameter for options of methods to use when constructing the confidence interval of
8		#' the proportion. Additionally, a stratification variable can be supplied via the `strata` element of the `variables`
9		#' argument.
10		#'
11		#' @inheritParams prop_strat_wilson
12		#' @inheritParams argument_convention
13		#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("estimate_proportion")`
14		#' to see available statistics for this function.
15		#' @param method (`string`)\cr the method used to construct the confidence interval
16		#' for proportion of successful outcomes; one of `waldcc`, `wald`, `clopper-pearson`,
17		#' `wilson`, `wilsonc`, `strat_wilson`, `strat_wilsonc`, `agresti-coull` or `jeffreys`.
18		#' @param long (`flag`)\cr whether a long description is required.
19		#'
20		#' @seealso [h_proportions]
21		#'
22		#' @name estimate_proportion
23		#' @order 1
24		NULL
25
26		#' @describeIn estimate_proportion Statistics function estimating a
27		#' proportion along with its confidence interval.
28		#'
29		#' @param df (`logical` or `data.frame`)\cr if only a logical vector is used,
30		#' it indicates whether each subject is a responder or not. `TRUE` represents
31		#' a successful outcome. If a `data.frame` is provided, also the `strata` variable
32		#' names must be provided in `variables` as a list element with the strata strings.
33		#' In the case of `data.frame`, the logical vector of responses must be indicated as a
34		#' variable name in `.var`.
35		#'
36		#' @return
37		#' * `s_proportion()` returns statistics `n_prop` (`n` and proportion) and `prop_ci` (proportion CI) for a
38		#' given variable.
39		#'
40		#' @examples
41		#' # Case with only logical vector.
42		#' rsp_v <- c(1, 0, 1, 0, 1, 1, 0, 0)
43		#' s_proportion(rsp_v)
44		#'
45		#' # Example for Stratified Wilson CI
46		#' nex <- 100 # Number of example rows
47		#' dta <- data.frame(
48		#' "rsp" = sample(c(TRUE, FALSE), nex, TRUE),
49		#' "grp" = sample(c("A", "B"), nex, TRUE),
50		#' "f1" = sample(c("a1", "a2"), nex, TRUE),
51		#' "f2" = sample(c("x", "y", "z"), nex, TRUE),
52		#' stringsAsFactors = TRUE
53		#' )
54		#'
55		#' s_proportion(
56		#' df = dta,
57		#' .var = "rsp",
58		#' variables = list(strata = c("f1", "f2")),
59		#' conf_level = 0.90,
60		#' method = "strat_wilson"
61		#' )
62		#'
63		#' @export
64		s_proportion <- function(df,
65		.var,
66		conf_level = 0.95,
67		method = c(
68		"waldcc", "wald", "clopper-pearson",
69		"wilson", "wilsonc", "strat_wilson", "strat_wilsonc",
70		"agresti-coull", "jeffreys"
71		),
72		weights = NULL,
73		max_iterations = 50,
74		variables = list(strata = NULL),
75		long = FALSE) {
76	167x	method <- match.arg(method)
77	167x	checkmate::assert_flag(long)
78	167x	assert_proportion_value(conf_level)
79
80	167x	if (!is.null(variables$strata)) {
81		# Checks for strata
82	!	if (missing(df)) stop("When doing stratified analysis a data.frame with specific columns is needed.")
83	!	strata_colnames <- variables$strata
84	!	checkmate::assert_character(strata_colnames, null.ok = FALSE)
85	!	strata_vars <- stats::setNames(as.list(strata_colnames), strata_colnames)
86	!	assert_df_with_variables(df, strata_vars)
87
88	!	strata <- interaction(df[strata_colnames])
89	!	strata <- as.factor(strata)
90
91		# Pushing down checks to prop_strat_wilson
92	167x	} else if (checkmate::test_subset(method, c("strat_wilson", "strat_wilsonc"))) {
93	!	stop("To use stratified methods you need to specify the strata variables.")
94		}
95	167x	if (checkmate::test_atomic_vector(df)) {
96	167x	rsp <- as.logical(df)
97		} else {
98	!	rsp <- as.logical(df[[.var]])
99		}
100	167x	n <- sum(rsp)
101	167x	p_hat <- mean(rsp)
102
103	167x	prop_ci <- switch(method,
104	167x	"clopper-pearson" = prop_clopper_pearson(rsp, conf_level),
105	167x	"wilson" = prop_wilson(rsp, conf_level),
106	167x	"wilsonc" = prop_wilson(rsp, conf_level, correct = TRUE),
107	167x	"strat_wilson" = prop_strat_wilson(rsp,
108	167x	strata,
109	167x	weights,
110	167x	conf_level,
111	167x	max_iterations,
112	167x	correct = FALSE
113	167x	)$conf_int,
114	167x	"strat_wilsonc" = prop_strat_wilson(rsp,
115	167x	strata,
116	167x	weights,
117	167x	conf_level,
118	167x	max_iterations,
119	167x	correct = TRUE
120	167x	)$conf_int,
121	167x	"wald" = prop_wald(rsp, conf_level),
122	167x	"waldcc" = prop_wald(rsp, conf_level, correct = TRUE),
123	167x	"agresti-coull" = prop_agresti_coull(rsp, conf_level),
124	167x	"jeffreys" = prop_jeffreys(rsp, conf_level)
125		)
126
127	167x	list(
128	167x	"n_prop" = formatters::with_label(c(n, p_hat), "Responders"),
129	167x	"prop_ci" = formatters::with_label(
130	167x	x = 100 * prop_ci, label = d_proportion(conf_level, method, long = long)
131		)
132		)
133		}
134
135		#' @describeIn estimate_proportion Formatted analysis function which is used as `afun`
136		#' in `estimate_proportion()`.
137		#'
138		#' @return
139		#' * `a_proportion()` returns the corresponding list with formatted [rtables::CellValue()].
140		#'
141		#' @export
142		a_proportion <- make_afun(
143		s_proportion,
144		.formats = c(n_prop = "xx (xx.x%)", prop_ci = "(xx.x, xx.x)")
145		)
146
147		#' @describeIn estimate_proportion Layout-creating function which can take statistics function arguments
148		#' and additional format arguments. This function is a wrapper for [rtables::analyze()].
149		#'
150		#' @return
151		#' * `estimate_proportion()` returns a layout object suitable for passing to further layouting functions,
152		#' or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
153		#' the statistics from `s_proportion()` to the table layout.
154		#'
155		#' @examples
156		#' dta_test <- data.frame(
157		#' USUBJID = paste0("S", 1:12),
158		#' ARM = rep(LETTERS[1:3], each = 4),
159		#' AVAL = rep(LETTERS[1:3], each = 4)
160		#' )
161		#'
162		#' basic_table() %>%
163		#' split_cols_by("ARM") %>%
164		#' estimate_proportion(vars = "AVAL") %>%
165		#' build_table(df = dta_test)
166		#'
167		#' @export
168		#' @order 2
169		estimate_proportion <- function(lyt,
170		vars,
171		conf_level = 0.95,
172		method = c(
173		"waldcc", "wald", "clopper-pearson",
174		"wilson", "wilsonc", "strat_wilson", "strat_wilsonc",
175		"agresti-coull", "jeffreys"
176		),
177		weights = NULL,
178		max_iterations = 50,
179		variables = list(strata = NULL),
180		long = FALSE,
181		na_str = default_na_str(),
182		nested = TRUE,
183		...,
184		show_labels = "hidden",
185		table_names = vars,
186		.stats = NULL,
187		.formats = NULL,
188		.labels = NULL,
189		.indent_mods = NULL) {
190	3x	extra_args <- list(
191	3x	conf_level = conf_level, method = method, weights = weights, max_iterations = max_iterations,
192	3x	variables = variables, long = long, ...
193		)
194
195	3x	afun <- make_afun(
196	3x	a_proportion,
197	3x	.stats = .stats,
198	3x	.formats = .formats,
199	3x	.labels = .labels,
200	3x	.indent_mods = .indent_mods
201		)
202	3x	analyze(
203	3x	lyt,
204	3x	vars,
205	3x	afun = afun,
206	3x	na_str = na_str,
207	3x	nested = nested,
208	3x	extra_args = extra_args,
209	3x	show_labels = show_labels,
210	3x	table_names = table_names
211		)
212		}
213
214		#' Helper functions for calculating proportion confidence intervals
215		#'
216		#' @description `r lifecycle::badge("stable")`
217		#'
218		#' Functions to calculate different proportion confidence intervals for use in [estimate_proportion()].
219		#'
220		#' @inheritParams argument_convention
221		#' @inheritParams estimate_proportion
222		#'
223		#' @return Confidence interval of a proportion.
224		#'
225		#' @seealso [estimate_proportion], descriptive function [d_proportion()],
226		#' and helper functions [strata_normal_quantile()] and [update_weights_strat_wilson()].
227		#'
228		#' @name h_proportions
229		NULL
230
231		#' @describeIn h_proportions Calculates the Wilson interval by calling [stats::prop.test()].
232		#' Also referred to as Wilson score interval.
233		#'
234		#' @examples
235		#' rsp <- c(
236		#' TRUE, TRUE, TRUE, TRUE, TRUE,
237		#' FALSE, FALSE, FALSE, FALSE, FALSE
238		#' )
239		#' prop_wilson(rsp, conf_level = 0.9)
240		#'
241		#' @export
242		prop_wilson <- function(rsp, conf_level, correct = FALSE) {
243	5x	y <- stats::prop.test(
244	5x	sum(rsp),
245	5x	length(rsp),
246	5x	correct = correct,
247	5x	conf.level = conf_level
248		)
249
250	5x	as.numeric(y$conf.int)
251		}
252
253		#' @describeIn h_proportions Calculates the stratified Wilson confidence
254		#' interval for unequal proportions as described in \insertCite{Yan2010-jt;textual}{tern}
255		#'
256		#' @param strata (`factor`)\cr variable with one level per stratum and same length as `rsp`.
257		#' @param weights (`numeric` or `NULL`)\cr weights for each level of the strata. If `NULL`, they are
258		#' estimated using the iterative algorithm proposed in \insertCite{Yan2010-jt;textual}{tern} that
259		#' minimizes the weighted squared length of the confidence interval.
260		#' @param max_iterations (`count`)\cr maximum number of iterations for the iterative procedure used
261		#' to find estimates of optimal weights.
262		#' @param correct (`flag`)\cr whether to include the continuity correction. For further information, see for example
263		#' for [stats::prop.test()].
264		#'
265		#' @references
266		#' \insertRef{Yan2010-jt}{tern}
267		#'
268		#' @examples
269		#' # Stratified Wilson confidence interval with unequal probabilities
270		#'
271		#' set.seed(1)
272		#' rsp <- sample(c(TRUE, FALSE), 100, TRUE)
273		#' strata_data <- data.frame(
274		#' "f1" = sample(c("a", "b"), 100, TRUE),
275		#' "f2" = sample(c("x", "y", "z"), 100, TRUE),
276		#' stringsAsFactors = TRUE
277		#' )
278		#' strata <- interaction(strata_data)
279		#' n_strata <- ncol(table(rsp, strata)) # Number of strata
280		#'
281		#' prop_strat_wilson(
282		#' rsp = rsp, strata = strata,
283		#' conf_level = 0.90
284		#' )
285		#'
286		#' # Not automatic setting of weights
287		#' prop_strat_wilson(
288		#' rsp = rsp, strata = strata,
289		#' weights = rep(1 / n_strata, n_strata),
290		#' conf_level = 0.90
291		#' )
292		#'
293		#' @export
294		prop_strat_wilson <- function(rsp,
295		strata,
296		weights = NULL,
297		conf_level = 0.95,
298		max_iterations = NULL,
299		correct = FALSE) {
300	20x	checkmate::assert_logical(rsp, any.missing = FALSE)
301	20x	checkmate::assert_factor(strata, len = length(rsp))
302	20x	assert_proportion_value(conf_level)
303
304	20x	tbl <- table(rsp, strata)
305	20x	n_strata <- length(unique(strata))
306
307		# Checking the weights and maximum number of iterations.
308	20x	do_iter <- FALSE
309	20x	if (is.null(weights)) {
310	6x	weights <- rep(1 / n_strata, n_strata) # Initialization for iterative procedure
311	6x	do_iter <- TRUE
312
313		# Iteration parameters
314	2x	if (is.null(max_iterations)) max_iterations <- 10
315	6x	checkmate::assert_int(max_iterations, na.ok = FALSE, null.ok = FALSE, lower = 1)
316		}
317	20x	checkmate::assert_numeric(weights, lower = 0, upper = 1, any.missing = FALSE, len = n_strata)
318	20x	sum_weights <- checkmate::assert_int(sum(weights))
319	!	if (as.integer(sum_weights + 0.5) != 1L) stop("Sum of weights must be 1L.")
320
321	20x	xs <- tbl["TRUE", ]
322	20x	ns <- colSums(tbl)
323	20x	use_stratum <- (ns > 0)
324	20x	ns <- ns[use_stratum]
325	20x	xs <- xs[use_stratum]
326	20x	ests <- xs / ns
327	20x	vars <- ests * (1 - ests) / ns
328
329	20x	strata_qnorm <- strata_normal_quantile(vars, weights, conf_level)
330
331		# Iterative setting of weights if they were not set externally
332	20x	weights_new <- if (do_iter) {
333	6x	update_weights_strat_wilson(vars, strata_qnorm, weights, ns, max_iterations, conf_level)$weights
334		} else {
335	14x	weights
336		}
337
338	20x	strata_conf_level <- 2 * stats::pnorm(strata_qnorm) - 1
339
340	20x	ci_by_strata <- Map(
341	20x	function(x, n) {
342		# Classic Wilson's confidence interval
343	139x	suppressWarnings(stats::prop.test(x, n, correct = correct, conf.level = strata_conf_level)$conf.int)
344		},
345	20x	x = xs,
346	20x	n = ns
347		)
348	20x	lower_by_strata <- sapply(ci_by_strata, "[", 1L)
349	20x	upper_by_strata <- sapply(ci_by_strata, "[", 2L)
350
351	20x	lower <- sum(weights_new * lower_by_strata)
352	20x	upper <- sum(weights_new * upper_by_strata)
353
354		# Return values
355	20x	if (do_iter) {
356	6x	list(
357	6x	conf_int = c(
358	6x	lower = lower,
359	6x	upper = upper
360		),
361	6x	weights = weights_new
362		)
363		} else {
364	14x	list(
365	14x	conf_int = c(
366	14x	lower = lower,
367	14x	upper = upper
368		)
369		)
370		}
371		}
372
373		#' @describeIn h_proportions Calculates the Clopper-Pearson interval by calling [stats::binom.test()].
374		#' Also referred to as the `exact` method.
375		#'
376		#' @examples
377		#' prop_clopper_pearson(rsp, conf_level = .95)
378		#'
379		#' @export
380		prop_clopper_pearson <- function(rsp,
381		conf_level) {
382	1x	y <- stats::binom.test(
383	1x	x = sum(rsp),
384	1x	n = length(rsp),
385	1x	conf.level = conf_level
386		)
387	1x	as.numeric(y$conf.int)
388		}
389
390		#' @describeIn h_proportions Calculates the Wald interval by following the usual textbook definition
391		#' for a single proportion confidence interval using the normal approximation.
392		#'
393		#' @param correct (`flag`)\cr whether to apply continuity correction.
394		#'
395		#' @examples
396		#' prop_wald(rsp, conf_level = 0.95)
397		#' prop_wald(rsp, conf_level = 0.95, correct = TRUE)
398		#'
399		#' @export
400		prop_wald <- function(rsp, conf_level, correct = FALSE) {
401	163x	n <- length(rsp)
402	163x	p_hat <- mean(rsp)
403	163x	z <- stats::qnorm((1 + conf_level) / 2)
404	163x	q_hat <- 1 - p_hat
405	163x	correct <- if (correct) 1 / (2 * n) else 0
406
407	163x	err <- z * sqrt(p_hat * q_hat) / sqrt(n) + correct
408	163x	l_ci <- max(0, p_hat - err)
409	163x	u_ci <- min(1, p_hat + err)
410
411	163x	c(l_ci, u_ci)
412		}
413
414		#' @describeIn h_proportions Calculates the Agresti-Coull interval. Constructed (for 95% CI) by adding two successes
415		#' and two failures to the data and then using the Wald formula to construct a CI.
416		#'
417		#' @examples
418		#' prop_agresti_coull(rsp, conf_level = 0.95)
419		#'
420		#' @export
421		prop_agresti_coull <- function(rsp, conf_level) {
422	3x	n <- length(rsp)
423	3x	x_sum <- sum(rsp)
424	3x	z <- stats::qnorm((1 + conf_level) / 2)
425
426		# Add here both z^2 / 2 successes and failures.
427	3x	x_sum_tilde <- x_sum + z^2 / 2
428	3x	n_tilde <- n + z^2
429
430		# Then proceed as with the Wald interval.
431	3x	p_tilde <- x_sum_tilde / n_tilde
432	3x	q_tilde <- 1 - p_tilde
433	3x	err <- z * sqrt(p_tilde * q_tilde) / sqrt(n_tilde)
434	3x	l_ci <- max(0, p_tilde - err)
435	3x	u_ci <- min(1, p_tilde + err)
436
437	3x	c(l_ci, u_ci)
438		}
439
440		#' @describeIn h_proportions Calculates the Jeffreys interval, an equal-tailed interval based on the
441		#' non-informative Jeffreys prior for a binomial proportion.
442		#'
443		#' @examples
444		#' prop_jeffreys(rsp, conf_level = 0.95)
445		#'
446		#' @export
447		prop_jeffreys <- function(rsp,
448		conf_level) {
449	5x	n <- length(rsp)
450	5x	x_sum <- sum(rsp)
451
452	5x	alpha <- 1 - conf_level
453	5x	l_ci <- ifelse(
454	5x	x_sum == 0,
455	5x	0,
456	5x	stats::qbeta(alpha / 2, x_sum + 0.5, n - x_sum + 0.5)
457		)
458
459	5x	u_ci <- ifelse(
460	5x	x_sum == n,
461	5x	1,
462	5x	stats::qbeta(1 - alpha / 2, x_sum + 0.5, n - x_sum + 0.5)
463		)
464
465	5x	c(l_ci, u_ci)
466		}
467
468		#' Description of the proportion summary
469		#'
470		#' @description `r lifecycle::badge("stable")`
471		#'
472		#' This is a helper function that describes the analysis in [s_proportion()].
473		#'
474		#' @inheritParams s_proportion
475		#' @param long (`flag`)\cr whether a long or a short (default) description is required.
476		#'
477		#' @return String describing the analysis.
478		#'
479		#' @export
480		d_proportion <- function(conf_level,
481		method,
482		long = FALSE) {
483	179x	label <- paste0(conf_level * 100, "% CI")
484
485	!	if (long) label <- paste(label, "for Response Rates")
486
487	179x	method_part <- switch(method,
488	179x	"clopper-pearson" = "Clopper-Pearson",
489	179x	"waldcc" = "Wald, with correction",
490	179x	"wald" = "Wald, without correction",
491	179x	"wilson" = "Wilson, without correction",
492	179x	"strat_wilson" = "Stratified Wilson, without correction",
493	179x	"wilsonc" = "Wilson, with correction",
494	179x	"strat_wilsonc" = "Stratified Wilson, with correction",
495	179x	"agresti-coull" = "Agresti-Coull",
496	179x	"jeffreys" = "Jeffreys",
497	179x	stop(paste(method, "does not have a description"))
498		)
499
500	179x	paste0(label, " (", method_part, ")")
501		}
502
503		#' Helper function for the estimation of stratified quantiles
504		#'
505		#' @description `r lifecycle::badge("stable")`
506		#'
507		#' This function wraps the estimation of stratified percentiles when we assume
508		#' the approximation for large numbers. This is necessary only in the case
509		#' proportions for each strata are unequal.
510		#'
511		#' @inheritParams argument_convention
512		#' @inheritParams prop_strat_wilson
513		#'
514		#' @return Stratified quantile.
515		#'
516		#' @seealso [prop_strat_wilson()]
517		#'
518		#' @examples
519		#' strata_data <- table(data.frame(
520		#' "f1" = sample(c(TRUE, FALSE), 100, TRUE),
521		#' "f2" = sample(c("x", "y", "z"), 100, TRUE),
522		#' stringsAsFactors = TRUE
523		#' ))
524		#' ns <- colSums(strata_data)
525		#' ests <- strata_data["TRUE", ] / ns
526		#' vars <- ests * (1 - ests) / ns
527		#' weights <- rep(1 / length(ns), length(ns))
528		#'
529		#' strata_normal_quantile(vars, weights, 0.95)
530		#'
531		#' @export
532		strata_normal_quantile <- function(vars, weights, conf_level) {
533	43x	summands <- weights^2 * vars
534		# Stratified quantile
535	43x	sqrt(sum(summands)) / sum(sqrt(summands)) * stats::qnorm((1 + conf_level) / 2)
536		}
537
538		#' Helper function for the estimation of weights for `prop_strat_wilson()`
539		#'
540		#' @description `r lifecycle::badge("stable")`
541		#'
542		#' This function wraps the iteration procedure that allows you to estimate
543		#' the weights for each proportional strata. This assumes to minimize the
544		#' weighted squared length of the confidence interval.
545		#'
546		#' @inheritParams prop_strat_wilson
547		#' @param vars (`numeric`)\cr normalized proportions for each strata.
548		#' @param strata_qnorm (`numeric(1)`)\cr initial estimation with identical weights of the quantiles.
549		#' @param initial_weights (`numeric`)\cr initial weights used to calculate `strata_qnorm`. This can
550		#' be optimized in the future if we need to estimate better initial weights.
551		#' @param n_per_strata (`numeric`)\cr number of elements in each strata.
552		#' @param max_iterations (`integer(1)`)\cr maximum number of iterations to be tried. Convergence is always checked.
553		#' @param tol (`numeric(1)`)\cr tolerance threshold for convergence.
554		#'
555		#' @return A `list` of 3 elements: `n_it`, `weights`, and `diff_v`.
556		#'
557		#' @seealso For references and details see [prop_strat_wilson()].
558		#'
559		#' @examples
560		#' vs <- c(0.011, 0.013, 0.012, 0.014, 0.017, 0.018)
561		#' sq <- 0.674
562		#' ws <- rep(1 / length(vs), length(vs))
563		#' ns <- c(22, 18, 17, 17, 14, 12)
564		#'
565		#' update_weights_strat_wilson(vs, sq, ws, ns, 100, 0.95, 0.001)
566		#'
567		#' @export
568		update_weights_strat_wilson <- function(vars,
569		strata_qnorm,
570		initial_weights,
571		n_per_strata,
572		max_iterations = 50,
573		conf_level = 0.95,
574		tol = 0.001) {
575	9x	it <- 0
576	9x	diff_v <- NULL
577
578	9x	while (it < max_iterations) {
579	21x	it <- it + 1
580	21x	weights_new_t <- (1 + strata_qnorm^2 / n_per_strata)^2
581	21x	weights_new_b <- (vars + strata_qnorm^2 / (4 * n_per_strata^2))
582	21x	weights_new <- weights_new_t / weights_new_b
583	21x	weights_new <- weights_new / sum(weights_new)
584	21x	strata_qnorm <- strata_normal_quantile(vars, weights_new, conf_level)
585	21x	diff_v <- c(diff_v, sum(abs(weights_new - initial_weights)))
586	8x	if (diff_v[length(diff_v)] < tol) break
587	13x	initial_weights <- weights_new
588		}
589
590	9x	if (it == max_iterations) {
591	1x	warning("The heuristic to find weights did not converge with max_iterations = ", max_iterations)
592		}
593
594	9x	list(
595	9x	"n_it" = it,
596	9x	"weights" = weights_new,
597	9x	"diff_v" = diff_v
598		)
599		}

1		# Utility functions to cooperate with {rtables} package
2
3		#' Convert table into matrix of strings
4		#'
5		#' @description `r lifecycle::badge("stable")`
6		#'
7		#' Helper function to use mostly within tests. `with_spaces`parameter allows
8		#' to test not only for content but also indentation and table structure.
9		#' `print_txt_to_copy` instead facilitate the testing development by returning a well
10		#' formatted text that needs only to be copied and pasted in the expected output.
11		#'
12		#' @inheritParams formatters::toString
13		#' @param x (`VTableTree`)\cr `rtables` table object.
14		#' @param with_spaces (`flag`)\cr whether the tested table should keep the indentation and other relevant spaces.
15		#' @param print_txt_to_copy (`flag`)\cr utility to have a way to copy the input table directly
16		#' into the expected variable instead of copying it too manually.
17		#'
18		#' @return A `matrix` of `string`s. If `print_txt_to_copy = TRUE` the well formatted printout of the
19		#' table will be printed to console, ready to be copied as a expected value.
20		#'
21		#' @examples
22		#' tbl <- basic_table() %>%
23		#' split_rows_by("SEX") %>%
24		#' split_cols_by("ARM") %>%
25		#' analyze("AGE") %>%
26		#' build_table(tern_ex_adsl)
27		#'
28		#' to_string_matrix(tbl, widths = ceiling(propose_column_widths(tbl) / 2))
29		#'
30		#' @export
31		to_string_matrix <- function(x, widths = NULL, max_width = NULL,
32		hsep = formatters::default_hsep(),
33		with_spaces = TRUE, print_txt_to_copy = FALSE) {
34	11x	checkmate::assert_flag(with_spaces)
35	11x	checkmate::assert_flag(print_txt_to_copy)
36	11x	checkmate::assert_int(max_width, null.ok = TRUE)
37
38	11x	if (inherits(x, "MatrixPrintForm")) {
39	!	tx <- x
40		} else {
41	11x	tx <- matrix_form(x, TRUE)
42		}
43
44	11x	tf_wrap <- FALSE
45	11x	if (!is.null(max_width)) {
46	!	tf_wrap <- TRUE
47		}
48
49		# Producing the matrix to test
50	11x	if (with_spaces) {
51	2x	out <- strsplit(toString(tx, widths = widths, tf_wrap = tf_wrap, max_width = max_width, hsep = hsep), "\n")[[1]]
52		} else {
53	9x	out <- tx$strings
54		}
55
56		# Printing to console formatted output that needs to be copied in "expected"
57	11x	if (print_txt_to_copy) {
58	2x	out_tmp <- out
59	2x	if (!with_spaces) {
60	1x	out_tmp <- apply(out, 1, paste0, collapse = '", "')
61		}
62	2x	cat(paste0('c(\n "', paste0(out_tmp, collapse = '",\n "'), '"\n)'))
63		}
64
65		# Return values
66	11x	return(out)
67		}
68
69		#' Blank for missing input
70		#'
71		#' Helper function to use in tabulating model results.
72		#'
73		#' @param x (`vector`)\cr input for a cell.
74		#'
75		#' @return An empty `character` vector if all entries in `x` are missing (`NA`), otherwise
76		#' the unlisted version of `x`.
77		#'
78		#' @keywords internal
79		unlist_and_blank_na <- function(x) {
80	267x	unl <- unlist(x)
81	267x	if (all(is.na(unl))) {
82	161x	character()
83		} else {
84	106x	unl
85		}
86		}
87
88		#' Constructor for content functions given a data frame with flag input
89		#'
90		#' This can be useful for tabulating model results.
91		#'
92		#' @param analysis_var (`string`)\cr variable name for the column containing values to be returned by the
93		#' content function.
94		#' @param flag_var (`string`)\cr variable name for the logical column identifying which row should be returned.
95		#' @param format (`string`)\cr `rtables` format to use.
96		#'
97		#' @return A content function which gives `df$analysis_var` at the row identified by
98		#' `.df_row$flag` in the given format.
99		#'
100		#' @keywords internal
101		cfun_by_flag <- function(analysis_var,
102		flag_var,
103		format = "xx",
104		.indent_mods = NULL) {
105	61x	checkmate::assert_string(analysis_var)
106	61x	checkmate::assert_string(flag_var)
107	61x	function(df, labelstr) {
108	265x	row_index <- which(df[[flag_var]])
109	265x	x <- unlist_and_blank_na(df[[analysis_var]][row_index])
110	265x	formatters::with_label(
111	265x	rcell(x, format = format, indent_mod = .indent_mods),
112	265x	labelstr
113		)
114		}
115		}
116
117		#' Content row function to add row total to labels
118		#'
119		#' This takes the label of the latest row split level and adds the row total from `df` in parentheses.
120		#' This function differs from [c_label_n_alt()] by taking row counts from `df` rather than
121		#' `alt_counts_df`, and is used by [add_rowcounts()] when `alt_counts` is set to `FALSE`.
122		#'
123		#' @inheritParams argument_convention
124		#'
125		#' @return A list with formatted [rtables::CellValue()] with the row count value and the correct label.
126		#'
127		#' @note It is important here to not use `df` but rather `.N_row` in the implementation, because
128		#' the former is already split by columns and will refer to the first column of the data only.
129		#'
130		#' @seealso [c_label_n_alt()] which performs the same function but retrieves row counts from
131		#' `alt_counts_df` instead of `df`.
132		#'
133		#' @keywords internal
134		c_label_n <- function(df,
135		labelstr,
136		.N_row) { # nolint
137	273x	label <- paste0(labelstr, " (N=", .N_row, ")")
138	273x	in_rows(
139	273x	.list = list(row_count = formatters::with_label(c(.N_row, .N_row), label)),
140	273x	.formats = c(row_count = function(x, ...) "")
141		)
142		}
143
144		#' Content row function to add `alt_counts_df` row total to labels
145		#'
146		#' This takes the label of the latest row split level and adds the row total from `alt_counts_df`
147		#' in parentheses. This function differs from [c_label_n()] by taking row counts from `alt_counts_df`
148		#' rather than `df`, and is used by [add_rowcounts()] when `alt_counts` is set to `TRUE`.
149		#'
150		#' @inheritParams argument_convention
151		#'
152		#' @return A list with formatted [rtables::CellValue()] with the row count value and the correct label.
153		#'
154		#' @seealso [c_label_n()] which performs the same function but retrieves row counts from `df` instead
155		#' of `alt_counts_df`.
156		#'
157		#' @keywords internal
158		c_label_n_alt <- function(df,
159		labelstr,
160		.alt_df_row) {
161	7x	N_row_alt <- nrow(.alt_df_row) # nolint
162	7x	label <- paste0(labelstr, " (N=", N_row_alt, ")")
163	7x	in_rows(
164	7x	.list = list(row_count = formatters::with_label(c(N_row_alt, N_row_alt), label)),
165	7x	.formats = c(row_count = function(x, ...) "")
166		)
167		}
168
169		#' Layout-creating function to add row total counts
170		#'
171		#' @description `r lifecycle::badge("stable")`
172		#'
173		#' This works analogously to [rtables::add_colcounts()] but on the rows. This function
174		#' is a wrapper for [rtables::summarize_row_groups()].
175		#'
176		#' @inheritParams argument_convention
177		#' @param alt_counts (`flag`)\cr whether row counts should be taken from `alt_counts_df` (`TRUE`)
178		#' or from `df` (`FALSE`). Defaults to `FALSE`.
179		#'
180		#' @return A modified layout where the latest row split labels now have the row-wise
181		#' total counts (i.e. without column-based subsetting) attached in parentheses.
182		#'
183		#' @note Row count values are contained in these row count rows but are not displayed
184		#' so that they are not considered zero rows by default when pruning.
185		#'
186		#' @examples
187		#' basic_table() %>%
188		#' split_cols_by("ARM") %>%
189		#' add_colcounts() %>%
190		#' split_rows_by("RACE", split_fun = drop_split_levels) %>%
191		#' add_rowcounts() %>%
192		#' analyze("AGE", afun = list_wrap_x(summary), format = "xx.xx") %>%
193		#' build_table(DM)
194		#'
195		#' @export
196		add_rowcounts <- function(lyt, alt_counts = FALSE) {
197	7x	summarize_row_groups(
198	7x	lyt,
199	7x	cfun = if (alt_counts) c_label_n_alt else c_label_n
200		)
201		}
202
203		#' Obtain column indices
204		#'
205		#' @description `r lifecycle::badge("stable")`
206		#'
207		#' Helper function to extract column indices from a `VTableTree` for a given
208		#' vector of column names.
209		#'
210		#' @param table_tree (`VTableTree`)\cr `rtables` table object to extract the indices from.
211		#' @param col_names (`character`)\cr vector of column names.
212		#'
213		#' @return A vector of column indices.
214		#'
215		#' @export
216		h_col_indices <- function(table_tree, col_names) {
217	1256x	checkmate::assert_class(table_tree, "VTableNodeInfo")
218	1256x	checkmate::assert_subset(col_names, names(attr(col_info(table_tree), "cextra_args")), empty.ok = FALSE)
219	1256x	match(col_names, names(attr(col_info(table_tree), "cextra_args")))
220		}
221
222		#' Labels or names of list elements
223		#'
224		#' Internal helper function for working with nested statistic function results which typically
225		#' don't have labels but names that we can use.
226		#'
227		#' @param x (`list`)\cr a list.
228		#'
229		#' @return A `character` vector with the labels or names for the list elements.
230		#'
231		#' @keywords internal
232		labels_or_names <- function(x) {
233	190x	checkmate::assert_multi_class(x, c("data.frame", "list"))
234	190x	labs <- sapply(x, obj_label)
235	190x	nams <- rlang::names2(x)
236	190x	label_is_null <- sapply(labs, is.null)
237	190x	result <- unlist(ifelse(label_is_null, nams, labs))
238	190x	return(result)
239		}
240
241		#' Convert to `rtable`
242		#'
243		#' @description `r lifecycle::badge("stable")`
244		#'
245		#' This is a new generic function to convert objects to `rtable` tables.
246		#'
247		#' @param x (`data.frame`)\cr the object which should be converted to an `rtable`.
248		#' @param ... additional arguments for methods.
249		#'
250		#' @return An `rtables` table object. Note that the concrete class will depend on the method used.
251		#'
252		#' @export
253		as.rtable <- function(x, ...) { # nolint
254	3x	UseMethod("as.rtable", x)
255		}
256
257		#' @describeIn as.rtable Method for converting a `data.frame` that contains numeric columns to `rtable`.
258		#'
259		#' @param format (`string` or `function`)\cr the format which should be used for the columns.
260		#'
261		#' @method as.rtable data.frame
262		#'
263		#' @examples
264		#' x <- data.frame(
265		#' a = 1:10,
266		#' b = rnorm(10)
267		#' )
268		#' as.rtable(x)
269		#'
270		#' @export
271		as.rtable.data.frame <- function(x, format = "xx.xx", ...) {
272	3x	checkmate::assert_numeric(unlist(x))
273	2x	do.call(
274	2x	rtable,
275	2x	c(
276	2x	list(
277	2x	header = labels_or_names(x),
278	2x	format = format
279		),
280	2x	Map(
281	2x	function(row, row_name) {
282	20x	do.call(
283	20x	rrow,
284	20x	c(as.list(unname(row)),
285	20x	row.name = row_name
286		)
287		)
288		},
289	2x	row = as.data.frame(t(x)),
290	2x	row_name = rownames(x)
291		)
292		)
293		)
294		}
295
296		#' Split parameters
297		#'
298		#' @description `r lifecycle::badge("stable")`
299		#'
300		#' It divides the data in the vector `param` into the groups defined by `f` based on specified `values`. It is relevant
301		#' in `rtables` layers so as to distribute parameters `.stats` or' `.formats` into lists with items corresponding to
302		#' specific analysis function.
303		#'
304		#' @param param (`vector`)\cr the parameter to be split.
305		#' @param value (`vector`)\cr the value used to split.
306		#' @param f (`list`)\cr the reference to make the split.
307		#'
308		#' @return A named `list` with the same element names as `f`, each containing the elements specified in `.stats`.
309		#'
310		#' @examples
311		#' f <- list(
312		#' surv = c("pt_at_risk", "event_free_rate", "rate_se", "rate_ci"),
313		#' surv_diff = c("rate_diff", "rate_diff_ci", "ztest_pval")
314		#' )
315		#'
316		#' .stats <- c("pt_at_risk", "rate_diff")
317		#' h_split_param(.stats, .stats, f = f)
318		#'
319		#' # $surv
320		#' # [1] "pt_at_risk"
321		#' #
322		#' # $surv_diff
323		#' # [1] "rate_diff"
324		#'
325		#' .formats <- c("pt_at_risk" = "xx", "event_free_rate" = "xxx")
326		#' h_split_param(.formats, names(.formats), f = f)
327		#'
328		#' # $surv
329		#' # pt_at_risk event_free_rate
330		#' # "xx" "xxx"
331		#' #
332		#' # $surv_diff
333		#' # NULL
334		#'
335		#' @export
336		h_split_param <- function(param,
337		value,
338		f) {
339	26x	y <- lapply(f, function(x) param[value %in% x])
340	26x	lapply(y, function(x) if (length(x) == 0) NULL else x)
341		}
342
343		#' Get selected statistics names
344		#'
345		#' Helper function to be used for creating `afun`.
346		#'
347		#' @param .stats (`vector` or `NULL`)\cr input to the layout creating function. Note that `NULL` means
348		#' in this context that all default statistics should be used.
349		#' @param all_stats (`character`)\cr all statistics which can be selected here potentially.
350		#'
351		#' @return A `character` vector with the selected statistics.
352		#'
353		#' @keywords internal
354		afun_selected_stats <- function(.stats, all_stats) {
355	2x	checkmate::assert_character(.stats, null.ok = TRUE)
356	2x	checkmate::assert_character(all_stats)
357	2x	if (is.null(.stats)) {
358	1x	all_stats
359		} else {
360	1x	intersect(.stats, all_stats)
361		}
362		}
363
364		#' Add variable labels to top left corner in table
365		#'
366		#' @description `r lifecycle::badge("stable")`
367		#'
368		#' Helper layout-creating function to append the variable labels of a given variables vector
369		#' from a given dataset in the top left corner. If a variable label is not found then the
370		#' variable name itself is used instead. Multiple variable labels are concatenated with slashes.
371		#'
372		#' @inheritParams argument_convention
373		#' @param vars (`character`)\cr variable names of which the labels are to be looked up in `df`.
374		#' @param indent (`integer(1)`)\cr non-negative number of nested indent space, default to 0L which means no indent.
375		#' 1L means two spaces indent, 2L means four spaces indent and so on.
376		#'
377		#' @return A modified layout with the new variable label(s) added to the top-left material.
378		#'
379		#' @note This is not an optimal implementation of course, since we are using here the data set
380		#' itself during the layout creation. When we have a more mature `rtables` implementation then
381		#' this will also be improved or not necessary anymore.
382		#'
383		#' @examples
384		#' lyt <- basic_table() %>%
385		#' split_cols_by("ARM") %>%
386		#' add_colcounts() %>%
387		#' split_rows_by("SEX") %>%
388		#' append_varlabels(DM, "SEX") %>%
389		#' analyze("AGE", afun = mean) %>%
390		#' append_varlabels(DM, "AGE", indent = 1)
391		#' build_table(lyt, DM)
392		#'
393		#' lyt <- basic_table() %>%
394		#' split_cols_by("ARM") %>%
395		#' split_rows_by("SEX") %>%
396		#' analyze("AGE", afun = mean) %>%
397		#' append_varlabels(DM, c("SEX", "AGE"))
398		#' build_table(lyt, DM)
399		#'
400		#' @export
401		append_varlabels <- function(lyt, df, vars, indent = 0L) {
402	3x	if (checkmate::test_flag(indent)) {
403	!	warning("indent argument is now accepting integers. Boolean indent will be converted to integers.")
404	!	indent <- as.integer(indent)
405		}
406
407	3x	checkmate::assert_data_frame(df)
408	3x	checkmate::assert_character(vars)
409	3x	checkmate::assert_count(indent)
410
411	3x	lab <- formatters::var_labels(df[vars], fill = TRUE)
412	3x	lab <- paste(lab, collapse = " / ")
413	3x	space <- paste(rep(" ", indent * 2), collapse = "")
414	3x	lab <- paste0(space, lab)
415
416	3x	append_topleft(lyt, lab)
417		}
418
419		#' Default string replacement for `NA` values
420		#'
421		#' @description `r lifecycle::badge("stable")`
422		#'
423		#' The default string used to represent `NA` values. This value is used as the default
424		#' value for the `na_str` argument throughout the `tern` package, and printed in place
425		#' of `NA` values in output tables. If not specified for each `tern` function by the user
426		#' via the `na_str` argument, or in the R environment options via [set_default_na_str()],
427		#' then `NA` is used.
428		#'
429		#' @param na_str (`string`)\cr single string value to set in the R environment options as
430		#' the default value to replace `NA`s. Use `getOption("tern_default_na_str")` to check the
431		#' current value set in the R environment (defaults to `NULL` if not set).
432		#'
433		#' @name default_na_str
434		NULL
435
436		#' @describeIn default_na_str Accessor for default `NA` value replacement string.
437		#'
438		#' @return
439		#' * `default_na_str` returns the current value if an R environment option has been set
440		#' for `"tern_default_na_str"`, or `NA_character_` otherwise.
441		#'
442		#' @examples
443		#' # Default settings
444		#' default_na_str()
445		#' getOption("tern_default_na_str")
446		#'
447		#' # Set custom value
448		#' set_default_na_str("<Missing>")
449		#'
450		#' # Settings after value has been set
451		#' default_na_str()
452		#' getOption("tern_default_na_str")
453		#'
454		#' @export
455		default_na_str <- function() {
456	330x	getOption("tern_default_na_str", default = NA_character_)
457		}
458
459		#' @describeIn default_na_str Setter for default `NA` value replacement string. Sets the
460		#' option `"tern_default_na_str"` within the R environment.
461		#'
462		#' @return
463		#' * `set_default_na_str` has no return value.
464		#'
465		#' @export
466		set_default_na_str <- function(na_str) {
467	3x	checkmate::assert_character(na_str, len = 1, null.ok = TRUE)
468	3x	options("tern_default_na_str" = na_str)
469		}

1		#' Helper functions for subgroup treatment effect pattern (STEP) calculations
2		#'
3		#' @description `r lifecycle::badge("stable")`
4		#'
5		#' Helper functions that are used internally for the STEP calculations.
6		#'
7		#' @inheritParams argument_convention
8		#'
9		#' @name h_step
10		#' @include control_step.R
11		NULL
12
13		#' @describeIn h_step Creates the windows for STEP, based on the control settings
14		#' provided.
15		#'
16		#' @param x (`numeric`)\cr biomarker value(s) to use (without `NA`).
17		#' @param control (named `list`)\cr output from `control_step()`.
18		#'
19		#' @return
20		#' * `h_step_window()` returns a list containing the window-selection matrix `sel`
21		#' and the interval information matrix `interval`.
22		#'
23		#' @export
24		h_step_window <- function(x,
25		control = control_step()) {
26	12x	checkmate::assert_numeric(x, min.len = 1, any.missing = FALSE)
27	12x	checkmate::assert_list(control, names = "named")
28
29	12x	sel <- matrix(FALSE, length(x), control$num_points)
30	12x	out <- matrix(0, control$num_points, 3)
31	12x	colnames(out) <- paste("Interval", c("Center", "Lower", "Upper"))
32	12x	if (control$use_percentile) {
33		# Create windows according to percentile cutoffs.
34	9x	out <- cbind(out, out)
35	9x	colnames(out)[1:3] <- paste("Percentile", c("Center", "Lower", "Upper"))
36	9x	xs <- seq(0, 1, length.out = control$num_points + 2)[-1]
37	9x	for (i in seq_len(control$num_points)) {
38	185x	out[i, 2:3] <- c(
39	185x	max(xs[i] - control$bandwidth, 0),
40	185x	min(xs[i] + control$bandwidth, 1)
41		)
42	185x	out[i, 5:6] <- stats::quantile(x, out[i, 2:3])
43	185x	sel[, i] <- x >= out[i, 5] & x <= out[i, 6]
44		}
45		# Center is the middle point of the percentile window.
46	9x	out[, 1] <- xs[-control$num_points - 1]
47	9x	out[, 4] <- stats::quantile(x, out[, 1])
48		} else {
49		# Create windows according to cutoffs.
50	3x	m <- c(min(x), max(x))
51	3x	xs <- seq(m[1], m[2], length.out = control$num_points + 2)[-1]
52	3x	for (i in seq_len(control$num_points)) {
53	11x	out[i, 2:3] <- c(
54	11x	max(xs[i] - control$bandwidth, m[1]),
55	11x	min(xs[i] + control$bandwidth, m[2])
56		)
57	11x	sel[, i] <- x >= out[i, 2] & x <= out[i, 3]
58		}
59		# Center is the same as the point for predicting.
60	3x	out[, 1] <- xs[-control$num_points - 1]
61		}
62	12x	list(sel = sel, interval = out)
63		}
64
65		#' @describeIn h_step Calculates the estimated treatment effect estimate
66		#' on the linear predictor scale and corresponding standard error from a STEP `model` fitted
67		#' on `data` given `variables` specification, for a single biomarker value `x`.
68		#' This works for both `coxph` and `glm` models, i.e. for calculating log hazard ratio or log odds
69		#' ratio estimates.
70		#'
71		#' @param model (`coxph` or `glm`)\cr the regression model object.
72		#'
73		#' @return
74		#' * `h_step_trt_effect()` returns a vector with elements `est` and `se`.
75		#'
76		#' @export
77		h_step_trt_effect <- function(data,
78		model,
79		variables,
80		x) {
81	208x	checkmate::assert_multi_class(model, c("coxph", "glm"))
82	208x	checkmate::assert_number(x)
83	208x	assert_df_with_variables(data, variables)
84	208x	checkmate::assert_factor(data[[variables$arm]], n.levels = 2)
85
86	208x	newdata <- data[c(1, 1), ]
87	208x	newdata[, variables$biomarker] <- x
88	208x	newdata[, variables$arm] <- levels(data[[variables$arm]])
89	208x	model_terms <- stats::delete.response(stats::terms(model))
90	208x	model_frame <- stats::model.frame(model_terms, data = newdata, xlev = model$xlevels)
91	208x	mat <- stats::model.matrix(model_terms, data = model_frame, contrasts.arg = model$contrasts)
92	208x	coefs <- stats::coef(model)
93		# Note: It is important to use the coef subset from matrix, otherwise intercept and
94		# strata are included for coxph() models.
95	208x	mat <- mat[, names(coefs)]
96	208x	mat_diff <- diff(mat)
97	208x	est <- mat_diff %*% coefs
98	208x	var <- mat_diff %% stats::vcov(model) %% t(mat_diff)
99	208x	se <- sqrt(var)
100	208x	c(
101	208x	est = est,
102	208x	se = se
103		)
104		}
105
106		#' @describeIn h_step Builds the model formula used in survival STEP calculations.
107		#'
108		#' @return
109		#' * `h_step_survival_formula()` returns a model formula.
110		#'
111		#' @export
112		h_step_survival_formula <- function(variables,
113		control = control_step()) {
114	10x	checkmate::assert_character(variables$covariates, null.ok = TRUE)
115
116	10x	assert_list_of_variables(variables[c("arm", "biomarker", "event", "time")])
117	10x	form <- paste0("Surv(", variables$time, ", ", variables$event, ") ~ ", variables$arm)
118	10x	if (control$degree > 0) {
119	5x	form <- paste0(form, " * stats::poly(", variables$biomarker, ", degree = ", control$degree, ", raw = TRUE)")
120		}
121	10x	if (!is.null(variables$covariates)) {
122	6x	form <- paste(form, "+", paste(variables$covariates, collapse = "+"))
123		}
124	10x	if (!is.null(variables$strata)) {
125	2x	form <- paste0(form, " + strata(", paste0(variables$strata, collapse = ", "), ")")
126		}
127	10x	stats::as.formula(form)
128		}
129
130		#' @describeIn h_step Estimates the model with `formula` built based on
131		#' `variables` in `data` for a given `subset` and `control` parameters for the
132		#' Cox regression.
133		#'
134		#' @param formula (`formula`)\cr the regression model formula.
135		#' @param subset (`logical`)\cr subset vector.
136		#'
137		#' @return
138		#' * `h_step_survival_est()` returns a matrix of number of observations `n`,
139		#' `events`, log hazard ratio estimates `loghr`, standard error `se`,
140		#' and Wald confidence interval bounds `ci_lower` and `ci_upper`. One row is
141		#' included for each biomarker value in `x`.
142		#'
143		#' @export
144		h_step_survival_est <- function(formula,
145		data,
146		variables,
147		x,
148		subset = rep(TRUE, nrow(data)),
149		control = control_coxph()) {
150	55x	checkmate::assert_formula(formula)
151	55x	assert_df_with_variables(data, variables)
152	55x	checkmate::assert_logical(subset, min.len = 1, any.missing = FALSE)
153	55x	checkmate::assert_numeric(x, min.len = 1, any.missing = FALSE)
154	55x	checkmate::assert_list(control, names = "named")
155
156		# Note: `subset` in `coxph` needs to be an expression referring to `data` variables.
157	55x	data$.subset <- subset
158	55x	coxph_warnings <- NULL
159	55x	tryCatch(
160	55x	withCallingHandlers(
161	55x	expr = {
162	55x	fit <- survival::coxph(
163	55x	formula = formula,
164	55x	data = data,
165	55x	subset = .subset,
166	55x	ties = control$ties
167		)
168		},
169	55x	warning = function(w) {
170	1x	coxph_warnings <<- c(coxph_warnings, w)
171	1x	invokeRestart("muffleWarning")
172		}
173		),
174	55x	finally = {
175		}
176		)
177	55x	if (!is.null(coxph_warnings)) {
178	1x	warning(paste(
179	1x	"Fit warnings occurred, please consider using a simpler model, or",
180	1x	"larger `bandwidth`, less `num_points` in `control_step()` settings"
181		))
182		}
183		# Produce a matrix with one row per `x` and columns `est` and `se`.
184	55x	estimates <- t(vapply(
185	55x	X = x,
186	55x	FUN = h_step_trt_effect,
187	55x	FUN.VALUE = c(1, 2),
188	55x	data = data,
189	55x	model = fit,
190	55x	variables = variables
191		))
192	55x	q_norm <- stats::qnorm((1 + control$conf_level) / 2)
193	55x	cbind(
194	55x	n = fit$n,
195	55x	events = fit$nevent,
196	55x	loghr = estimates[, "est"],
197	55x	se = estimates[, "se"],
198	55x	ci_lower = estimates[, "est"] - q_norm * estimates[, "se"],
199	55x	ci_upper = estimates[, "est"] + q_norm * estimates[, "se"]
200		)
201		}
202
203		#' @describeIn h_step Builds the model formula used in response STEP calculations.
204		#'
205		#' @return
206		#' * `h_step_rsp_formula()` returns a model formula.
207		#'
208		#' @export
209		h_step_rsp_formula <- function(variables,
210		control = c(control_step(), control_logistic())) {
211	14x	checkmate::assert_character(variables$covariates, null.ok = TRUE)
212	14x	assert_list_of_variables(variables[c("arm", "biomarker", "response")])
213	14x	response_definition <- sub(
214	14x	pattern = "response",
215	14x	replacement = variables$response,
216	14x	x = control$response_definition,
217	14x	fixed = TRUE
218		)
219	14x	form <- paste0(response_definition, " ~ ", variables$arm)
220	14x	if (control$degree > 0) {
221	8x	form <- paste0(form, " * stats::poly(", variables$biomarker, ", degree = ", control$degree, ", raw = TRUE)")
222		}
223	14x	if (!is.null(variables$covariates)) {
224	8x	form <- paste(form, "+", paste(variables$covariates, collapse = "+"))
225		}
226	14x	if (!is.null(variables$strata)) {
227	5x	strata_arg <- if (length(variables$strata) > 1) {
228	2x	paste0("I(interaction(", paste0(variables$strata, collapse = ", "), "))")
229		} else {
230	3x	variables$strata
231		}
232	5x	form <- paste0(form, "+ strata(", strata_arg, ")")
233		}
234	14x	stats::as.formula(form)
235		}
236
237		#' @describeIn h_step Estimates the model with `formula` built based on
238		#' `variables` in `data` for a given `subset` and `control` parameters for the
239		#' logistic regression.
240		#'
241		#' @param formula (`formula`)\cr the regression model formula.
242		#' @param subset (`logical`)\cr subset vector.
243		#'
244		#' @return
245		#' * `h_step_rsp_est()` returns a matrix of number of observations `n`, log odds
246		#' ratio estimates `logor`, standard error `se`, and Wald confidence interval bounds
247		#' `ci_lower` and `ci_upper`. One row is included for each biomarker value in `x`.
248		#'
249		#' @export
250		h_step_rsp_est <- function(formula,
251		data,
252		variables,
253		x,
254		subset = rep(TRUE, nrow(data)),
255		control = control_logistic()) {
256	58x	checkmate::assert_formula(formula)
257	58x	assert_df_with_variables(data, variables)
258	58x	checkmate::assert_logical(subset, min.len = 1, any.missing = FALSE)
259	58x	checkmate::assert_numeric(x, min.len = 1, any.missing = FALSE)
260	58x	checkmate::assert_list(control, names = "named")
261		# Note: `subset` in `glm` needs to be an expression referring to `data` variables.
262	58x	data$.subset <- subset
263	58x	fit_warnings <- NULL
264	58x	tryCatch(
265	58x	withCallingHandlers(
266	58x	expr = {
267	58x	fit <- if (is.null(variables$strata)) {
268	54x	stats::glm(
269	54x	formula = formula,
270	54x	data = data,
271	54x	subset = .subset,
272	54x	family = stats::binomial("logit")
273		)
274		} else {
275		# clogit needs coxph and strata imported
276	4x	survival::clogit(
277	4x	formula = formula,
278	4x	data = data,
279	4x	subset = .subset
280		)
281		}
282		},
283	58x	warning = function(w) {
284	19x	fit_warnings <<- c(fit_warnings, w)
285	19x	invokeRestart("muffleWarning")
286		}
287		),
288	58x	finally = {
289		}
290		)
291	58x	if (!is.null(fit_warnings)) {
292	13x	warning(paste(
293	13x	"Fit warnings occurred, please consider using a simpler model, or",
294	13x	"larger `bandwidth`, less `num_points` in `control_step()` settings"
295		))
296		}
297		# Produce a matrix with one row per `x` and columns `est` and `se`.
298	58x	estimates <- t(vapply(
299	58x	X = x,
300	58x	FUN = h_step_trt_effect,
301	58x	FUN.VALUE = c(1, 2),
302	58x	data = data,
303	58x	model = fit,
304	58x	variables = variables
305		))
306	58x	q_norm <- stats::qnorm((1 + control$conf_level) / 2)
307	58x	cbind(
308	58x	n = length(fit$y),
309	58x	logor = estimates[, "est"],
310	58x	se = estimates[, "se"],
311	58x	ci_lower = estimates[, "est"] - q_norm * estimates[, "se"],
312	58x	ci_upper = estimates[, "est"] + q_norm * estimates[, "se"]
313		)
314		}

1		#' Tabulate survival duration by subgroup
2		#'
3		#' @description `r lifecycle::badge("stable")`
4		#'
5		#' The [tabulate_survival_subgroups()] function creates a layout element to tabulate survival duration by subgroup,
6		#' returning statistics including median survival time and hazard ratio for each population subgroup. The table is
7		#' created from `df`, a list of data frames returned by [extract_survival_subgroups()], with the statistics to include
8		#' specified via the `vars` parameter.
9		#'
10		#' A forest plot can be created from the resulting table using the [g_forest()] function.
11		#'
12		#' @inheritParams argument_convention
13		#' @inheritParams survival_coxph_pairwise
14		#' @param df (`list`)\cr list of data frames containing all analysis variables. List should be
15		#' created using [extract_survival_subgroups()].
16		#' @param vars (`character`)\cr the names of statistics to be reported among:
17		#' * `n_tot_events`: Total number of events per group.
18		#' * `n_events`: Number of events per group.
19		#' * `n_tot`: Total number of observations per group.
20		#' * `n`: Number of observations per group.
21		#' * `median`: Median survival time.
22		#' * `hr`: Hazard ratio.
23		#' * `ci`: Confidence interval of hazard ratio.
24		#' * `pval`: p-value of the effect.
25		#' Note, one of the statistics `n_tot` and `n_tot_events`, as well as both `hr` and `ci`
26		#' are required.
27		#' @param time_unit (`string`)\cr label with unit of median survival time. Default `NULL` skips displaying unit.
28		#'
29		#' @details These functions create a layout starting from a data frame which contains
30		#' the required statistics. Tables typically used as part of forest plot.
31		#'
32		#' @seealso [extract_survival_subgroups()]
33		#'
34		#' @examples
35		#' library(dplyr)
36		#'
37		#' adtte <- tern_ex_adtte
38		#'
39		#' # Save variable labels before data processing steps.
40		#' adtte_labels <- formatters::var_labels(adtte)
41		#'
42		#' adtte_f <- adtte %>%
43		#' filter(
44		#' PARAMCD == "OS",
45		#' ARM %in% c("B: Placebo", "A: Drug X"),
46		#' SEX %in% c("M", "F")
47		#' ) %>%
48		#' mutate(
49		#' # Reorder levels of ARM to display reference arm before treatment arm.
50		#' ARM = droplevels(forcats::fct_relevel(ARM, "B: Placebo")),
51		#' SEX = droplevels(SEX),
52		#' AVALU = as.character(AVALU),
53		#' is_event = CNSR == 0
54		#' )
55		#' labels <- c(
56		#' "ARM" = adtte_labels[["ARM"]],
57		#' "SEX" = adtte_labels[["SEX"]],
58		#' "AVALU" = adtte_labels[["AVALU"]],
59		#' "is_event" = "Event Flag"
60		#' )
61		#' formatters::var_labels(adtte_f)[names(labels)] <- labels
62		#'
63		#' df <- extract_survival_subgroups(
64		#' variables = list(
65		#' tte = "AVAL",
66		#' is_event = "is_event",
67		#' arm = "ARM", subgroups = c("SEX", "BMRKR2")
68		#' ),
69		#' label_all = "Total Patients",
70		#' data = adtte_f
71		#' )
72		#' df
73		#'
74		#' df_grouped <- extract_survival_subgroups(
75		#' variables = list(
76		#' tte = "AVAL",
77		#' is_event = "is_event",
78		#' arm = "ARM", subgroups = c("SEX", "BMRKR2")
79		#' ),
80		#' data = adtte_f,
81		#' groups_lists = list(
82		#' BMRKR2 = list(
83		#' "low" = "LOW",
84		#' "low/medium" = c("LOW", "MEDIUM"),
85		#' "low/medium/high" = c("LOW", "MEDIUM", "HIGH")
86		#' )
87		#' )
88		#' )
89		#' df_grouped
90		#'
91		#' @name survival_duration_subgroups
92		#' @order 1
93		NULL
94
95		#' Prepare survival data for population subgroups in data frames
96		#'
97		#' @description `r lifecycle::badge("stable")`
98		#'
99		#' Prepares estimates of median survival times and treatment hazard ratios for population subgroups in
100		#' data frames. Simple wrapper for [h_survtime_subgroups_df()] and [h_coxph_subgroups_df()]. Result is a `list`
101		#' of two `data.frame`s: `survtime` and `hr`. `variables` corresponds to the names of variables found in `data`,
102		#' passed as a named `list` and requires elements `tte`, `is_event`, `arm` and optionally `subgroups` and `strata`.
103		#' `groups_lists` optionally specifies groupings for `subgroups` variables.
104		#'
105		#' @inheritParams argument_convention
106		#' @inheritParams survival_duration_subgroups
107		#' @inheritParams survival_coxph_pairwise
108		#'
109		#' @return A named `list` of two elements:
110		#' * `survtime`: A `data.frame` containing columns `arm`, `n`, `n_events`, `median`, `subgroup`, `var`,
111		#' `var_label`, and `row_type`.
112		#' * `hr`: A `data.frame` containing columns `arm`, `n_tot`, `n_tot_events`, `hr`, `lcl`, `ucl`, `conf_level`,
113		#' `pval`, `pval_label`, `subgroup`, `var`, `var_label`, and `row_type`.
114		#'
115		#' @seealso [survival_duration_subgroups]
116		#'
117		#' @export
118		extract_survival_subgroups <- function(variables,
119		data,
120		groups_lists = list(),
121		control = control_coxph(),
122		label_all = "All Patients") {
123	12x	if ("strat" %in% names(variables)) {
124	!	warning(
125	!	"Warning: the `strat` element name of the `variables` list argument to `extract_survival_subgroups() ",
126	!	"was deprecated in tern 0.9.4.\n ",
127	!	"Please use the name `strata` instead of `strat` in the `variables` argument."
128		)
129	!	variables[["strata"]] <- variables[["strat"]]
130		}
131
132	12x	df_survtime <- h_survtime_subgroups_df(
133	12x	variables,
134	12x	data,
135	12x	groups_lists = groups_lists,
136	12x	label_all = label_all
137		)
138	12x	df_hr <- h_coxph_subgroups_df(
139	12x	variables,
140	12x	data,
141	12x	groups_lists = groups_lists,
142	12x	control = control,
143	12x	label_all = label_all
144		)
145
146	12x	list(survtime = df_survtime, hr = df_hr)
147		}
148
149		#' @describeIn survival_duration_subgroups Formatted analysis function which is used as
150		#' `afun` in `tabulate_survival_subgroups()`.
151		#'
152		#' @return
153		#' * `a_survival_subgroups()` returns the corresponding list with formatted [rtables::CellValue()].
154		#'
155		#' @keywords internal
156		a_survival_subgroups <- function(.formats = list( # nolint start
157		n = "xx",
158		n_events = "xx",
159		n_tot_events = "xx",
160		median = "xx.x",
161		n_tot = "xx",
162		hr = list(format_extreme_values(2L)),
163		ci = list(format_extreme_values_ci(2L)),
164		pval = "x.xxxx \| (<0.0001)"
165		),
166		na_str = default_na_str()) { # nolint end
167	21x	checkmate::assert_list(.formats)
168	21x	checkmate::assert_subset(
169	21x	names(.formats),
170	21x	c("n", "n_events", "median", "n_tot", "n_tot_events", "hr", "ci", "pval", "riskdiff")
171		)
172
173	21x	afun_lst <- Map(
174	21x	function(stat, fmt, na_str) {
175	160x	function(df, labelstr = "", ...) {
176	312x	in_rows(
177	312x	.list = as.list(df[[stat]]),
178	312x	.labels = as.character(df$subgroup),
179	312x	.formats = fmt,
180	312x	.format_na_strs = na_str
181		)
182		}
183		},
184	21x	stat = names(.formats),
185	21x	fmt = .formats,
186	21x	na_str = na_str
187		)
188
189	21x	afun_lst
190		}
191
192		#' @describeIn survival_duration_subgroups Table-creating function which creates a table
193		#' summarizing survival by subgroup. This function is a wrapper for [rtables::analyze_colvars()]
194		#' and [rtables::summarize_row_groups()].
195		#'
196		#' @param label_all `r lifecycle::badge("deprecated")`\cr please assign the `label_all` parameter within the
197		#' [extract_survival_subgroups()] function when creating `df`.
198		#' @param riskdiff (`list`)\cr if a risk (proportion) difference column should be added, a list of settings to apply
199		#' within the column. See [control_riskdiff()] for details. If `NULL`, no risk difference column will be added. If
200		#' `riskdiff$arm_x` and `riskdiff$arm_y` are `NULL`, the first level of `df$survtime$arm` will be used as `arm_x`
201		#' and the second level as `arm_y`.
202		#'
203		#' @return An `rtables` table summarizing survival by subgroup.
204		#'
205		#' @examples
206		#' ## Table with default columns.
207		#' basic_table() %>%
208		#' tabulate_survival_subgroups(df, time_unit = adtte_f$AVALU[1])
209		#'
210		#' ## Table with a manually chosen set of columns: adding "pval".
211		#' basic_table() %>%
212		#' tabulate_survival_subgroups(
213		#' df = df,
214		#' vars = c("n_tot_events", "n_events", "median", "hr", "ci", "pval"),
215		#' time_unit = adtte_f$AVALU[1]
216		#' )
217		#'
218		#' @export
219		#' @order 2
220		tabulate_survival_subgroups <- function(lyt,
221		df,
222		vars = c("n_tot_events", "n_events", "median", "hr", "ci"),
223		groups_lists = list(),
224		label_all = lifecycle::deprecated(),
225		time_unit = NULL,
226		riskdiff = NULL,
227		na_str = default_na_str(),
228		.formats = c(
229		n = "xx", n_events = "xx", n_tot_events = "xx", median = "xx.x", n_tot = "xx",
230		hr = list(format_extreme_values(2L)), ci = list(format_extreme_values_ci(2L)),
231		pval = "x.xxxx \| (<0.0001)"
232		)) {
233	10x	checkmate::assert_list(riskdiff, null.ok = TRUE)
234	10x	checkmate::assert_true(any(c("n_tot", "n_tot_events") %in% vars))
235	10x	checkmate::assert_true(all(c("hr", "ci") %in% vars))
236
237	10x	if (lifecycle::is_present(label_all)) {
238	1x	lifecycle::deprecate_warn(
239	1x	"0.9.5", "tabulate_survival_subgroups(label_all)",
240	1x	details =
241	1x	"Please assign the `label_all` parameter within the `extract_survival_subgroups()` function when creating `df`."
242		)
243		}
244
245		# Create "ci" column from "lcl" and "ucl"
246	10x	df$hr$ci <- combine_vectors(df$hr$lcl, df$hr$ucl)
247
248		# Fill in missing formats with defaults
249	10x	default_fmts <- eval(formals(tabulate_survival_subgroups)$.formats)
250	10x	.formats <- c(.formats, default_fmts[vars[!vars %in% names(.formats)]])
251
252		# Extract additional parameters from df
253	10x	conf_level <- df$hr$conf_level[1]
254	10x	method <- df$hr$pval_label[1]
255	10x	colvars <- d_survival_subgroups_colvars(vars, conf_level = conf_level, method = method, time_unit = time_unit)
256	10x	survtime_vars <- intersect(colvars$vars, c("n", "n_events", "median"))
257	10x	hr_vars <- intersect(names(colvars$labels), c("n_tot", "n_tot_events", "hr", "ci", "pval"))
258	10x	colvars_survtime <- list(vars = survtime_vars, labels = colvars$labels[survtime_vars])
259	10x	colvars_hr <- list(vars = hr_vars, labels = colvars$labels[hr_vars])
260
261	10x	extra_args <- list(groups_lists = groups_lists, conf_level = conf_level, method = method)
262
263		# Get analysis function for each statistic
264	10x	afun_lst <- a_survival_subgroups(.formats = c(.formats, riskdiff = riskdiff$format), na_str = na_str)
265
266		# Add risk difference column
267	10x	if (!is.null(riskdiff)) {
268	1x	if (is.null(riskdiff$arm_x)) riskdiff$arm_x <- levels(df$survtime$arm)[1]
269	1x	if (is.null(riskdiff$arm_y)) riskdiff$arm_y <- levels(df$survtime$arm)[2]
270	1x	colvars_hr$vars <- c(colvars_hr$vars, "riskdiff")
271	1x	colvars_hr$labels <- c(colvars_hr$labels, riskdiff = riskdiff$col_label)
272	1x	arm_cols <- paste(rep(c("n_events", "n_events", "n", "n")), c(riskdiff$arm_x, riskdiff$arm_y), sep = "_")
273
274	1x	df_prop_diff <- df$survtime %>%
275	1x	dplyr::select(-"median") %>%
276	1x	tidyr::pivot_wider(
277	1x	id_cols = c("subgroup", "var", "var_label", "row_type"),
278	1x	names_from = "arm",
279	1x	values_from = c("n", "n_events")
280		) %>%
281	1x	dplyr::rowwise() %>%
282	1x	dplyr::mutate(
283	1x	riskdiff = stat_propdiff_ci(
284	1x	x = as.list(.data[[arm_cols[1]]]),
285	1x	y = as.list(.data[[arm_cols[2]]]),
286	1x	N_x = .data[[arm_cols[3]]],
287	1x	N_y = .data[[arm_cols[4]]]
288		)
289		) %>%
290	1x	dplyr::select(-dplyr::all_of(arm_cols))
291
292	1x	df$hr <- df$hr %>%
293	1x	dplyr::left_join(
294	1x	df_prop_diff,
295	1x	by = c("subgroup", "var", "var_label", "row_type")
296		)
297		}
298
299		# Add columns from table_survtime (optional)
300	10x	if (length(colvars_survtime$vars) > 0) {
301	9x	lyt_survtime <- split_cols_by(lyt = lyt, var = "arm")
302	9x	lyt_survtime <- split_rows_by(
303	9x	lyt = lyt_survtime,
304	9x	var = "row_type",
305	9x	split_fun = keep_split_levels("content"),
306	9x	nested = FALSE
307		)
308
309		# Add "All Patients" row
310	9x	lyt_survtime <- summarize_row_groups(
311	9x	lyt = lyt_survtime,
312	9x	var = "var_label",
313	9x	cfun = afun_lst[names(colvars_survtime$labels)],
314	9x	na_str = na_str,
315	9x	extra_args = extra_args
316		)
317	9x	lyt_survtime <- split_cols_by_multivar(
318	9x	lyt = lyt_survtime,
319	9x	vars = colvars_survtime$vars,
320	9x	varlabels = colvars_survtime$labels
321		)
322
323		# Add analysis rows
324	9x	if ("analysis" %in% df$survtime$row_type) {
325	8x	lyt_survtime <- split_rows_by(
326	8x	lyt = lyt_survtime,
327	8x	var = "row_type",
328	8x	split_fun = keep_split_levels("analysis"),
329	8x	nested = FALSE,
330	8x	child_labels = "hidden"
331		)
332	8x	lyt_survtime <- split_rows_by(lyt = lyt_survtime, var = "var_label", nested = TRUE)
333	8x	lyt_survtime <- analyze_colvars(
334	8x	lyt = lyt_survtime,
335	8x	afun = afun_lst[names(colvars_survtime$labels)],
336	8x	na_str = na_str,
337	8x	inclNAs = TRUE,
338	8x	extra_args = extra_args
339		)
340		}
341
342	9x	table_survtime <- build_table(lyt_survtime, df = df$survtime)
343		} else {
344	1x	table_survtime <- NULL
345		}
346
347		# Add columns from table_hr ("n_tot_events" or "n_tot", "or" and "ci" required)
348	10x	lyt_hr <- split_cols_by(lyt = lyt, var = "arm")
349	10x	lyt_hr <- split_rows_by(
350	10x	lyt = lyt_hr,
351	10x	var = "row_type",
352	10x	split_fun = keep_split_levels("content"),
353	10x	nested = FALSE
354		)
355	10x	lyt_hr <- summarize_row_groups(
356	10x	lyt = lyt_hr,
357	10x	var = "var_label",
358	10x	cfun = afun_lst[names(colvars_hr$labels)],
359	10x	na_str = na_str,
360	10x	extra_args = extra_args
361		)
362	10x	lyt_hr <- split_cols_by_multivar(
363	10x	lyt = lyt_hr,
364	10x	vars = colvars_hr$vars,
365	10x	varlabels = colvars_hr$labels
366		) %>%
367	10x	append_topleft("Baseline Risk Factors")
368
369		# Add analysis rows
370	10x	if ("analysis" %in% df$survtime$row_type) {
371	9x	lyt_hr <- split_rows_by(
372	9x	lyt = lyt_hr,
373	9x	var = "row_type",
374	9x	split_fun = keep_split_levels("analysis"),
375	9x	nested = FALSE,
376	9x	child_labels = "hidden"
377		)
378	9x	lyt_hr <- split_rows_by(lyt = lyt_hr, var = "var_label", nested = TRUE)
379	9x	lyt_hr <- analyze_colvars(
380	9x	lyt = lyt_hr,
381	9x	afun = afun_lst[names(colvars_hr$labels)],
382	9x	na_str = na_str,
383	9x	inclNAs = TRUE,
384	9x	extra_args = extra_args
385		)
386		}
387
388	10x	table_hr <- build_table(lyt_hr, df = df$hr)
389
390		# Join tables, add forest plot attributes
391	10x	n_tot_ids <- grep("^n_tot", colvars_hr$vars)
392	10x	if (is.null(table_survtime)) {
393	1x	result <- table_hr
394	1x	hr_id <- match("hr", colvars_hr$vars)
395	1x	ci_id <- match("ci", colvars_hr$vars)
396		} else {
397	9x	result <- cbind_rtables(table_hr[, n_tot_ids], table_survtime, table_hr[, -n_tot_ids])
398	9x	hr_id <- length(n_tot_ids) + ncol(table_survtime) + match("hr", colvars_hr$vars[-n_tot_ids])
399	9x	ci_id <- length(n_tot_ids) + ncol(table_survtime) + match("ci", colvars_hr$vars[-n_tot_ids])
400	9x	n_tot_ids <- seq_along(n_tot_ids)
401		}
402	10x	structure(
403	10x	result,
404	10x	forest_header = paste0(rev(levels(df$survtime$arm)), "\nBetter"),
405	10x	col_x = hr_id,
406	10x	col_ci = ci_id,
407	10x	col_symbol_size = n_tot_ids[1] # for scaling the symbol sizes in forest plots
408		)
409		}
410
411		#' Labels for column variables in survival duration by subgroup table
412		#'
413		#' @description `r lifecycle::badge("stable")`
414		#'
415		#' Internal function to check variables included in [tabulate_survival_subgroups()] and create column labels.
416		#'
417		#' @inheritParams tabulate_survival_subgroups
418		#' @inheritParams argument_convention
419		#' @param method (`string`)\cr p-value method for testing hazard ratio = 1.
420		#'
421		#' @return A `list` of variables and their labels to tabulate.
422		#'
423		#' @note At least one of `n_tot` and `n_tot_events` must be provided in `vars`.
424		#'
425		#' @export
426		d_survival_subgroups_colvars <- function(vars,
427		conf_level,
428		method,
429		time_unit = NULL) {
430	21x	checkmate::assert_character(vars)
431	21x	checkmate::assert_string(time_unit, null.ok = TRUE)
432	21x	checkmate::assert_subset(c("hr", "ci"), vars)
433	21x	checkmate::assert_true(any(c("n_tot", "n_tot_events") %in% vars))
434	21x	checkmate::assert_subset(
435	21x	vars,
436	21x	c("n", "n_events", "median", "n_tot", "n_tot_events", "hr", "ci", "pval")
437		)
438
439	21x	propcase_time_label <- if (!is.null(time_unit)) {
440	20x	paste0("Median (", time_unit, ")")
441		} else {
442	1x	"Median"
443		}
444
445	21x	varlabels <- c(
446	21x	n = "n",
447	21x	n_events = "Events",
448	21x	median = propcase_time_label,
449	21x	n_tot = "Total n",
450	21x	n_tot_events = "Total Events",
451	21x	hr = "Hazard Ratio",
452	21x	ci = paste0(100 * conf_level, "% Wald CI"),
453	21x	pval = method
454		)
455
456	21x	colvars <- vars
457
458		# The `lcl` variable is just a placeholder available in the analysis data,
459		# it is not acutally used in the tabulation.
460		# Variables used in the tabulation are lcl and ucl, see `a_survival_subgroups` for details.
461	21x	colvars[colvars == "ci"] <- "lcl"
462
463	21x	list(
464	21x	vars = colvars,
465	21x	labels = varlabels[vars]
466		)
467		}

1		#' Helper functions for incidence rate
2		#'
3		#' @description `r lifecycle::badge("stable")`
4		#'
5		#' @param control (`list`)\cr parameters for estimation details, specified by using
6		#' the helper function [control_incidence_rate()]. Possible parameter options are:
7		#' * `conf_level`: (`proportion`)\cr confidence level for the estimated incidence rate.
8		#' * `conf_type`: (`string`)\cr `normal` (default), `normal_log`, `exact`, or `byar`
9		#' for confidence interval type.
10		#' * `input_time_unit`: (`string`)\cr `day`, `week`, `month`, or `year` (default)
11		#' indicating time unit for data input.
12		#' * `num_pt_year`: (`numeric`)\cr time unit for desired output (in person-years).
13		#' @param person_years (`numeric(1)`)\cr total person-years at risk.
14		#' @param alpha (`numeric(1)`)\cr two-sided alpha-level for confidence interval.
15		#' @param n_events (`integer(1)`)\cr number of events observed.
16		#'
17		#' @return Estimated incidence rate, `rate`, and associated confidence interval, `rate_ci`.
18		#'
19		#' @seealso [incidence_rate]
20		#'
21		#' @name h_incidence_rate
22		NULL
23
24		#' @describeIn h_incidence_rate Helper function to estimate the incidence rate and
25		#' associated confidence interval.
26		#'
27		#' @keywords internal
28		h_incidence_rate <- function(person_years,
29		n_events,
30		control = control_incidence_rate()) {
31	18x	alpha <- 1 - control$conf_level
32	18x	est <- switch(control$conf_type,
33	18x	normal = h_incidence_rate_normal(person_years, n_events, alpha),
34	18x	normal_log = h_incidence_rate_normal_log(person_years, n_events, alpha),
35	18x	exact = h_incidence_rate_exact(person_years, n_events, alpha),
36	18x	byar = h_incidence_rate_byar(person_years, n_events, alpha)
37		)
38
39	18x	num_pt_year <- control$num_pt_year
40	18x	list(
41	18x	rate = est$rate * num_pt_year,
42	18x	rate_ci = est$rate_ci * num_pt_year
43		)
44		}
45
46		#' @describeIn h_incidence_rate Helper function to estimate the incidence rate and
47		#' associated confidence interval based on the normal approximation for the
48		#' incidence rate. Unit is one person-year.
49		#'
50		#' @examples
51		#' h_incidence_rate_normal(200, 2)
52		#'
53		#' @export
54		h_incidence_rate_normal <- function(person_years,
55		n_events,
56		alpha = 0.05) {
57	14x	checkmate::assert_number(person_years)
58	14x	checkmate::assert_number(n_events)
59	14x	assert_proportion_value(alpha)
60
61	14x	est <- n_events / person_years
62	14x	se <- sqrt(est / person_years)
63	14x	ci <- est + c(-1, 1) * stats::qnorm(1 - alpha / 2) * se
64
65	14x	list(rate = est, rate_ci = ci)
66		}
67
68		#' @describeIn h_incidence_rate Helper function to estimate the incidence rate and
69		#' associated confidence interval based on the normal approximation for the
70		#' logarithm of the incidence rate. Unit is one person-year.
71		#'
72		#' @examples
73		#' h_incidence_rate_normal_log(200, 2)
74		#'
75		#' @export
76		h_incidence_rate_normal_log <- function(person_years,
77		n_events,
78		alpha = 0.05) {
79	6x	checkmate::assert_number(person_years)
80	6x	checkmate::assert_number(n_events)
81	6x	assert_proportion_value(alpha)
82
83	6x	rate_est <- n_events / person_years
84	6x	rate_se <- sqrt(rate_est / person_years)
85	6x	lrate_est <- log(rate_est)
86	6x	lrate_se <- rate_se / rate_est
87	6x	ci <- exp(lrate_est + c(-1, 1) * stats::qnorm(1 - alpha / 2) * lrate_se)
88
89	6x	list(rate = rate_est, rate_ci = ci)
90		}
91
92		#' @describeIn h_incidence_rate Helper function to estimate the incidence rate and
93		#' associated exact confidence interval. Unit is one person-year.
94		#'
95		#' @examples
96		#' h_incidence_rate_exact(200, 2)
97		#'
98		#' @export
99		h_incidence_rate_exact <- function(person_years,
100		n_events,
101		alpha = 0.05) {
102	1x	checkmate::assert_number(person_years)
103	1x	checkmate::assert_number(n_events)
104	1x	assert_proportion_value(alpha)
105
106	1x	est <- n_events / person_years
107	1x	lcl <- stats::qchisq(p = (alpha) / 2, df = 2 * n_events) / (2 * person_years)
108	1x	ucl <- stats::qchisq(p = 1 - (alpha) / 2, df = 2 * n_events + 2) / (2 * person_years)
109
110	1x	list(rate = est, rate_ci = c(lcl, ucl))
111		}
112
113		#' @describeIn h_incidence_rate Helper function to estimate the incidence rate and
114		#' associated Byar's confidence interval. Unit is one person-year.
115		#'
116		#' @examples
117		#' h_incidence_rate_byar(200, 2)
118		#'
119		#' @export
120		h_incidence_rate_byar <- function(person_years,
121		n_events,
122		alpha = 0.05) {
123	1x	checkmate::assert_number(person_years)
124	1x	checkmate::assert_number(n_events)
125	1x	assert_proportion_value(alpha)
126
127	1x	est <- n_events / person_years
128	1x	seg_1 <- n_events + 0.5
129	1x	seg_2 <- 1 - 1 / (9 * (n_events + 0.5))
130	1x	seg_3 <- stats::qnorm(1 - alpha / 2) * sqrt(1 / (n_events + 0.5)) / 3
131	1x	lcl <- seg_1 * ((seg_2 - seg_3)^3) / person_years
132	1x	ucl <- seg_1 * ((seg_2 + seg_3)^3) / person_years
133
134	1x	list(rate = est, rate_ci = c(lcl, ucl))
135		}

1		#' Individual patient plots
2		#'
3		#' @description `r lifecycle::badge("stable")`
4		#'
5		#' Line plot(s) displaying trend in patients' parameter values over time is rendered.
6		#' Patients' individual baseline values can be added to the plot(s) as reference.
7		#'
8		#' @inheritParams argument_convention
9		#' @param xvar (`string`)\cr time point variable to be plotted on x-axis.
10		#' @param yvar (`string`)\cr continuous analysis variable to be plotted on y-axis.
11		#' @param xlab (`string`)\cr plot label for x-axis.
12		#' @param ylab (`string`)\cr plot label for y-axis.
13		#' @param id_var (`string`)\cr variable used as patient identifier.
14		#' @param title (`string`)\cr title for plot.
15		#' @param subtitle (`string`)\cr subtitle for plot.
16		#' @param add_baseline_hline (`flag`)\cr adds horizontal line at baseline y-value on
17		#' plot when `TRUE`.
18		#' @param yvar_baseline (`string`)\cr variable with baseline values only.
19		#' Ignored when `add_baseline_hline` is `FALSE`.
20		#' @param ggtheme (`theme`)\cr optional graphical theme function as provided
21		#' by `ggplot2` to control outlook of plot. Use `ggplot2::theme()` to tweak the display.
22		#' @param plotting_choices (`string`)\cr specifies options for displaying
23		#' plots. Must be one of `"all_in_one"`, `"split_by_max_obs"`, or `"separate_by_obs"`.
24		#' @param max_obs_per_plot (`integer(1)`)\cr number of observations to be plotted on one
25		#' plot. Ignored if `plotting_choices` is not `"separate_by_obs"`.
26		#' @param caption (`string`)\cr optional caption below the plot.
27		#' @param col (`character`)\cr line colors.
28		#'
29		#' @seealso Relevant helper function [h_g_ipp()].
30		#'
31		#' @name g_ipp
32		#' @aliases individual_patient_plot
33		NULL
34
35		#' Helper function to create simple line plot over time
36		#'
37		#' @description `r lifecycle::badge("stable")`
38		#'
39		#' Function that generates a simple line plot displaying parameter trends over time.
40		#'
41		#' @inheritParams argument_convention
42		#' @inheritParams g_ipp
43		#'
44		#' @return A `ggplot` line plot.
45		#'
46		#' @seealso [g_ipp()] which uses this function.
47		#'
48		#' @examples
49		#' library(dplyr)
50		#' library(nestcolor)
51		#'
52		#' # Select a small sample of data to plot.
53		#' adlb <- tern_ex_adlb %>%
54		#' filter(PARAMCD == "ALT", !(AVISIT %in% c("SCREENING", "BASELINE"))) %>%
55		#' slice(1:36)
56		#'
57		#' p <- h_g_ipp(
58		#' df = adlb,
59		#' xvar = "AVISIT",
60		#' yvar = "AVAL",
61		#' xlab = "Visit",
62		#' id_var = "USUBJID",
63		#' ylab = "SGOT/ALT (U/L)",
64		#' add_baseline_hline = TRUE
65		#' )
66		#' p
67		#'
68		#' @export
69		h_g_ipp <- function(df,
70		xvar,
71		yvar,
72		xlab,
73		ylab,
74		id_var,
75		title = "Individual Patient Plots",
76		subtitle = "",
77		caption = NULL,
78		add_baseline_hline = FALSE,
79		yvar_baseline = "BASE",
80		ggtheme = nestcolor::theme_nest(),
81		col = NULL) {
82	13x	checkmate::assert_string(xvar)
83	13x	checkmate::assert_string(yvar)
84	13x	checkmate::assert_string(yvar_baseline)
85	13x	checkmate::assert_string(id_var)
86	13x	checkmate::assert_string(xlab)
87	13x	checkmate::assert_string(ylab)
88	13x	checkmate::assert_string(title)
89	13x	checkmate::assert_string(subtitle)
90	13x	checkmate::assert_subset(c(xvar, yvar, yvar_baseline, id_var), colnames(df))
91	13x	checkmate::assert_data_frame(df)
92	13x	checkmate::assert_flag(add_baseline_hline)
93	13x	checkmate::assert_character(col, null.ok = TRUE)
94
95	13x	p <- ggplot2::ggplot(
96	13x	data = df,
97	13x	mapping = ggplot2::aes(
98	13x	x = .data[[xvar]],
99	13x	y = .data[[yvar]],
100	13x	group = .data[[id_var]],
101	13x	colour = .data[[id_var]]
102		)
103		) +
104	13x	ggplot2::geom_line(linewidth = 0.4) +
105	13x	ggplot2::geom_point(size = 2) +
106	13x	ggplot2::labs(
107	13x	x = xlab,
108	13x	y = ylab,
109	13x	title = title,
110	13x	subtitle = subtitle,
111	13x	caption = caption
112		) +
113	13x	ggtheme
114
115	13x	if (add_baseline_hline) {
116	12x	baseline_df <- df[, c(id_var, yvar_baseline)]
117	12x	baseline_df <- unique(baseline_df)
118
119	12x	p <- p +
120	12x	ggplot2::geom_hline(
121	12x	data = baseline_df,
122	12x	mapping = ggplot2::aes(
123	12x	yintercept = .data[[yvar_baseline]],
124	12x	colour = .data[[id_var]]
125		),
126	12x	linetype = "dotdash",
127	12x	linewidth = 0.4
128		) +
129	12x	ggplot2::geom_text(
130	12x	data = baseline_df,
131	12x	mapping = ggplot2::aes(
132	12x	x = 1,
133	12x	y = .data[[yvar_baseline]],
134	12x	label = .data[[id_var]],
135	12x	colour = .data[[id_var]]
136		),
137	12x	nudge_y = 0.025 * (max(df[, yvar], na.rm = TRUE) - min(df[, yvar], na.rm = TRUE)),
138	12x	vjust = "right",
139	12x	size = 2
140		)
141
142	12x	if (!is.null(col)) {
143	1x	p <- p +
144	1x	ggplot2::scale_color_manual(values = col)
145		}
146		}
147	13x	p
148		}
149
150		#' @describeIn g_ipp Plotting function for individual patient plots which, depending on user
151		#' preference, renders a single graphic or compiles a list of graphics that show trends in individual's parameter
152		#' values over time.
153		#'
154		#' @return A `ggplot` object or a list of `ggplot` objects.
155		#'
156		#' @examples
157		#' library(dplyr)
158		#' library(nestcolor)
159		#'
160		#' # Select a small sample of data to plot.
161		#' adlb <- tern_ex_adlb %>%
162		#' filter(PARAMCD == "ALT", !(AVISIT %in% c("SCREENING", "BASELINE"))) %>%
163		#' slice(1:36)
164		#'
165		#' plot_list <- g_ipp(
166		#' df = adlb,
167		#' xvar = "AVISIT",
168		#' yvar = "AVAL",
169		#' xlab = "Visit",
170		#' ylab = "SGOT/ALT (U/L)",
171		#' title = "Individual Patient Plots",
172		#' add_baseline_hline = TRUE,
173		#' plotting_choices = "split_by_max_obs",
174		#' max_obs_per_plot = 5
175		#' )
176		#' plot_list
177		#'
178		#' @export
179		g_ipp <- function(df,
180		xvar,
181		yvar,
182		xlab,
183		ylab,
184		id_var = "USUBJID",
185		title = "Individual Patient Plots",
186		subtitle = "",
187		caption = NULL,
188		add_baseline_hline = FALSE,
189		yvar_baseline = "BASE",
190		ggtheme = nestcolor::theme_nest(),
191		plotting_choices = c("all_in_one", "split_by_max_obs", "separate_by_obs"),
192		max_obs_per_plot = 4,
193		col = NULL) {
194	3x	checkmate::assert_count(max_obs_per_plot)
195	3x	checkmate::assert_subset(plotting_choices, c("all_in_one", "split_by_max_obs", "separate_by_obs"))
196	3x	checkmate::assert_character(col, null.ok = TRUE)
197
198	3x	plotting_choices <- match.arg(plotting_choices)
199
200	3x	if (plotting_choices == "all_in_one") {
201	1x	p <- h_g_ipp(
202	1x	df = df,
203	1x	xvar = xvar,
204	1x	yvar = yvar,
205	1x	xlab = xlab,
206	1x	ylab = ylab,
207	1x	id_var = id_var,
208	1x	title = title,
209	1x	subtitle = subtitle,
210	1x	caption = caption,
211	1x	add_baseline_hline = add_baseline_hline,
212	1x	yvar_baseline = yvar_baseline,
213	1x	ggtheme = ggtheme,
214	1x	col = col
215		)
216
217	1x	return(p)
218	2x	} else if (plotting_choices == "split_by_max_obs") {
219	1x	id_vec <- unique(df[[id_var]])
220	1x	id_list <- split(
221	1x	id_vec,
222	1x	rep(1:ceiling(length(id_vec) / max_obs_per_plot),
223	1x	each = max_obs_per_plot,
224	1x	length.out = length(id_vec)
225		)
226		)
227
228	1x	df_list <- list()
229	1x	plot_list <- list()
230
231	1x	for (i in seq_along(id_list)) {
232	2x	df_list[[i]] <- df[df[[id_var]] %in% id_list[[i]], ]
233
234	2x	plots <- h_g_ipp(
235	2x	df = df_list[[i]],
236	2x	xvar = xvar,
237	2x	yvar = yvar,
238	2x	xlab = xlab,
239	2x	ylab = ylab,
240	2x	id_var = id_var,
241	2x	title = title,
242	2x	subtitle = subtitle,
243	2x	caption = caption,
244	2x	add_baseline_hline = add_baseline_hline,
245	2x	yvar_baseline = yvar_baseline,
246	2x	ggtheme = ggtheme,
247	2x	col = col
248		)
249
250	2x	plot_list[[i]] <- plots
251		}
252	1x	return(plot_list)
253		} else {
254	1x	ind_df <- split(df, df[[id_var]])
255	1x	plot_list <- lapply(
256	1x	ind_df,
257	1x	function(x) {
258	8x	h_g_ipp(
259	8x	df = x,
260	8x	xvar = xvar,
261	8x	yvar = yvar,
262	8x	xlab = xlab,
263	8x	ylab = ylab,
264	8x	id_var = id_var,
265	8x	title = title,
266	8x	subtitle = subtitle,
267	8x	caption = caption,
268	8x	add_baseline_hline = add_baseline_hline,
269	8x	yvar_baseline = yvar_baseline,
270	8x	ggtheme = ggtheme,
271	8x	col = col
272		)
273		}
274		)
275
276	1x	return(plot_list)
277		}
278		}

1		#' Difference test for two proportions
2		#'
3		#' @description `r lifecycle::badge("stable")`
4		#'
5		#' The analyze function [test_proportion_diff()] creates a layout element to test the difference between two
6		#' proportions. The primary analysis variable, `vars`, indicates whether a response has occurred for each record. See
7		#' the `method` parameter for options of methods to use to calculate the p-value. Additionally, a stratification
8		#' variable can be supplied via the `strata` element of the `variables` argument.
9		#'
10		#' @inheritParams argument_convention
11		#' @param method (`string`)\cr one of `chisq`, `cmh`, `fisher`, or `schouten`; specifies the test used
12		#' to calculate the p-value.
13		#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("test_proportion_diff")`
14		#' to see available statistics for this function.
15		#'
16		#' @seealso [h_prop_diff_test]
17		#'
18		#' @name prop_diff_test
19		#' @order 1
20		NULL
21
22		#' @describeIn prop_diff_test Statistics function which tests the difference between two proportions.
23		#'
24		#' @return
25		#' * `s_test_proportion_diff()` returns a named `list` with a single item `pval` with an attribute `label`
26		#' describing the method used. The p-value tests the null hypothesis that proportions in two groups are the same.
27		#'
28		#' @keywords internal
29		s_test_proportion_diff <- function(df,
30		.var,
31		.ref_group,
32		.in_ref_col,
33		variables = list(strata = NULL),
34		method = c("chisq", "schouten", "fisher", "cmh")) {
35	45x	method <- match.arg(method)
36	45x	y <- list(pval = "")
37
38	45x	if (!.in_ref_col) {
39	45x	assert_df_with_variables(df, list(rsp = .var))
40	45x	assert_df_with_variables(.ref_group, list(rsp = .var))
41	45x	rsp <- factor(
42	45x	c(.ref_group[[.var]], df[[.var]]),
43	45x	levels = c("TRUE", "FALSE")
44		)
45	45x	grp <- factor(
46	45x	rep(c("ref", "Not-ref"), c(nrow(.ref_group), nrow(df))),
47	45x	levels = c("ref", "Not-ref")
48		)
49
50	45x	if (!is.null(variables$strata) \|\| method == "cmh") {
51	12x	strata <- variables$strata
52	12x	checkmate::assert_false(is.null(strata))
53	12x	strata_vars <- stats::setNames(as.list(strata), strata)
54	12x	assert_df_with_variables(df, strata_vars)
55	12x	assert_df_with_variables(.ref_group, strata_vars)
56	12x	strata <- c(interaction(.ref_group[strata]), interaction(df[strata]))
57		}
58
59	45x	tbl <- switch(method,
60	45x	cmh = table(grp, rsp, strata),
61	45x	table(grp, rsp)
62		)
63
64	45x	y$pval <- switch(method,
65	45x	chisq = prop_chisq(tbl),
66	45x	cmh = prop_cmh(tbl),
67	45x	fisher = prop_fisher(tbl),
68	45x	schouten = prop_schouten(tbl)
69		)
70		}
71
72	45x	y$pval <- formatters::with_label(y$pval, d_test_proportion_diff(method))
73	45x	y
74		}
75
76		#' Description of the difference test between two proportions
77		#'
78		#' @description `r lifecycle::badge("stable")`
79		#'
80		#' This is an auxiliary function that describes the analysis in `s_test_proportion_diff`.
81		#'
82		#' @inheritParams s_test_proportion_diff
83		#'
84		#' @return A `string` describing the test from which the p-value is derived.
85		#'
86		#' @export
87		d_test_proportion_diff <- function(method) {
88	59x	checkmate::assert_string(method)
89	59x	meth_part <- switch(method,
90	59x	"schouten" = "Chi-Squared Test with Schouten Correction",
91	59x	"chisq" = "Chi-Squared Test",
92	59x	"cmh" = "Cochran-Mantel-Haenszel Test",
93	59x	"fisher" = "Fisher's Exact Test",
94	59x	stop(paste(method, "does not have a description"))
95		)
96	59x	paste0("p-value (", meth_part, ")")
97		}
98
99		#' @describeIn prop_diff_test Formatted analysis function which is used as `afun` in `test_proportion_diff()`.
100		#'
101		#' @return
102		#' * `a_test_proportion_diff()` returns the corresponding list with formatted [rtables::CellValue()].
103		#'
104		#' @keywords internal
105		a_test_proportion_diff <- make_afun(
106		s_test_proportion_diff,
107		.formats = c(pval = "x.xxxx \| (<0.0001)"),
108		.indent_mods = c(pval = 1L)
109		)
110
111		#' @describeIn prop_diff_test Layout-creating function which can take statistics function arguments
112		#' and additional format arguments. This function is a wrapper for [rtables::analyze()].
113		#'
114		#' @return
115		#' * `test_proportion_diff()` returns a layout object suitable for passing to further layouting functions,
116		#' or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
117		#' the statistics from `s_test_proportion_diff()` to the table layout.
118		#'
119		#' @examples
120		#' dta <- data.frame(
121		#' rsp = sample(c(TRUE, FALSE), 100, TRUE),
122		#' grp = factor(rep(c("A", "B"), each = 50)),
123		#' strata = factor(rep(c("V", "W", "X", "Y", "Z"), each = 20))
124		#' )
125		#'
126		#' # With `rtables` pipelines.
127		#' l <- basic_table() %>%
128		#' split_cols_by(var = "grp", ref_group = "B") %>%
129		#' test_proportion_diff(
130		#' vars = "rsp",
131		#' method = "cmh", variables = list(strata = "strata")
132		#' )
133		#'
134		#' build_table(l, df = dta)
135		#'
136		#' @export
137		#' @order 2
138		test_proportion_diff <- function(lyt,
139		vars,
140		variables = list(strata = NULL),
141		method = c("chisq", "schouten", "fisher", "cmh"),
142		na_str = default_na_str(),
143		nested = TRUE,
144		...,
145		var_labels = vars,
146		show_labels = "hidden",
147		table_names = vars,
148		.stats = NULL,
149		.formats = NULL,
150		.labels = NULL,
151		.indent_mods = NULL) {
152	6x	extra_args <- list(variables = variables, method = method, ...)
153
154	6x	afun <- make_afun(
155	6x	a_test_proportion_diff,
156	6x	.stats = .stats,
157	6x	.formats = .formats,
158	6x	.labels = .labels,
159	6x	.indent_mods = .indent_mods
160		)
161	6x	analyze(
162	6x	lyt,
163	6x	vars,
164	6x	afun = afun,
165	6x	var_labels = var_labels,
166	6x	na_str = na_str,
167	6x	nested = nested,
168	6x	extra_args = extra_args,
169	6x	show_labels = show_labels,
170	6x	table_names = table_names
171		)
172		}
173
174		#' Helper functions to test proportion differences
175		#'
176		#' Helper functions to implement various tests on the difference between two proportions.
177		#'
178		#' @param tbl (`matrix`)\cr matrix with two groups in rows and the binary response (`TRUE`/`FALSE`) in columns.
179		#'
180		#' @return A p-value.
181		#'
182		#' @seealso [prop_diff_test()] for implementation of these helper functions.
183		#'
184		#' @name h_prop_diff_test
185		NULL
186
187		#' @describeIn h_prop_diff_test Performs Chi-Squared test. Internally calls [stats::prop.test()].
188		#'
189		#' @keywords internal
190		prop_chisq <- function(tbl) {
191	41x	checkmate::assert_integer(c(ncol(tbl), nrow(tbl)), lower = 2, upper = 2)
192	41x	tbl <- tbl[, c("TRUE", "FALSE")]
193	41x	if (any(colSums(tbl) == 0)) {
194	2x	return(1)
195		}
196	39x	stats::prop.test(tbl, correct = FALSE)$p.value
197		}
198
199		#' @describeIn h_prop_diff_test Performs stratified Cochran-Mantel-Haenszel test. Internally calls
200		#' [stats::mantelhaen.test()]. Note that strata with less than two observations are automatically discarded.
201		#'
202		#' @param ary (`array`, 3 dimensions)\cr array with two groups in rows, the binary response
203		#' (`TRUE`/`FALSE`) in columns, and the strata in the third dimension.
204		#'
205		#' @keywords internal
206		prop_cmh <- function(ary) {
207	16x	checkmate::assert_array(ary)
208	16x	checkmate::assert_integer(c(ncol(ary), nrow(ary)), lower = 2, upper = 2)
209	16x	checkmate::assert_integer(length(dim(ary)), lower = 3, upper = 3)
210	16x	strata_sizes <- apply(ary, MARGIN = 3, sum)
211	16x	if (any(strata_sizes < 5)) {
212	1x	warning("<5 data points in some strata. CMH test may be incorrect.")
213	1x	ary <- ary[, , strata_sizes > 1]
214		}
215
216	16x	stats::mantelhaen.test(ary, correct = FALSE)$p.value
217		}
218
219		#' @describeIn h_prop_diff_test Performs the Chi-Squared test with Schouten correction.
220		#'
221		#' @seealso Schouten correction is based upon \insertCite{Schouten1980-kd;textual}{tern}.
222		#'
223		#' @keywords internal
224		prop_schouten <- function(tbl) {
225	100x	checkmate::assert_integer(c(ncol(tbl), nrow(tbl)), lower = 2, upper = 2)
226	100x	tbl <- tbl[, c("TRUE", "FALSE")]
227	100x	if (any(colSums(tbl) == 0)) {
228	1x	return(1)
229		}
230
231	99x	n <- sum(tbl)
232	99x	n1 <- sum(tbl[1, ])
233	99x	n2 <- sum(tbl[2, ])
234
235	99x	ad <- diag(tbl)
236	99x	bc <- diag(apply(tbl, 2, rev))
237	99x	ac <- tbl[, 1]
238	99x	bd <- tbl[, 2]
239
240	99x	t_schouten <- (n - 1) *
241	99x	(abs(prod(ad) - prod(bc)) - 0.5 * min(n1, n2))^2 /
242	99x	(n1 * n2 * sum(ac) * sum(bd))
243
244	99x	1 - stats::pchisq(t_schouten, df = 1)
245		}
246
247		#' @describeIn h_prop_diff_test Performs the Fisher's exact test. Internally calls [stats::fisher.test()].
248		#'
249		#' @keywords internal
250		prop_fisher <- function(tbl) {
251	2x	checkmate::assert_integer(c(ncol(tbl), nrow(tbl)), lower = 2, upper = 2)
252	2x	tbl <- tbl[, c("TRUE", "FALSE")]
253	2x	stats::fisher.test(tbl)$p.value
254		}

1		#' Count occurrences by grade
2		#'
3		#' @description `r lifecycle::badge("stable")`
4		#'
5		#' The analyze function [count_occurrences_by_grade()] creates a layout element to calculate occurrence counts by grade.
6		#'
7		#' This function analyzes primary analysis variable `var` which indicates toxicity grades. The `id` variable
8		#' is used to indicate unique subject identifiers (defaults to `USUBJID`). The user can also supply a list of
9		#' custom groups of grades to analyze via the `grade_groups` parameter. The `remove_single` argument will
10		#' remove single grades from the analysis so that only grade groups are analyzed.
11		#'
12		#' If there are multiple grades recorded for one patient only the highest grade level is counted.
13		#'
14		#' The summarize function [summarize_occurrences_by_grade()] performs the same function as
15		#' [count_occurrences_by_grade()] except it creates content rows, not data rows, to summarize the current table
16		#' row/column context and operates on the level of the latest row split or the root of the table if no row splits have
17		#' occurred.
18		#'
19		#' @inheritParams argument_convention
20		#' @param grade_groups (named `list` of `character`)\cr list containing groupings of grades.
21		#' @param remove_single (`flag`)\cr `TRUE` to not include the elements of one-element grade groups
22		#' in the the output list; in this case only the grade groups names will be included in the output. If
23		#' `only_grade_groups` is set to `TRUE` this argument is ignored.
24		#' @param only_grade_groups (`flag`)\cr whether only the specified grade groups should be
25		#' included, with individual grade rows removed (`TRUE`), or all grades and grade groups
26		#' should be displayed (`FALSE`).
27		#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("count_occurrences_by_grade")`
28		#' to see available statistics for this function.
29		#'
30		#' @seealso Relevant helper function [h_append_grade_groups()].
31		#'
32		#' @name count_occurrences_by_grade
33		#' @order 1
34		NULL
35
36		#' Helper function for `s_count_occurrences_by_grade()`
37		#'
38		#' @description `r lifecycle::badge("stable")`
39		#'
40		#' Helper function for [s_count_occurrences_by_grade()] to insert grade groupings into list with
41		#' individual grade frequencies. The order of the final result follows the order of `grade_groups`.
42		#' The elements under any-grade group (if any), i.e. the grade group equal to `refs` will be moved to
43		#' the end. Grade groups names must be unique.
44		#'
45		#' @inheritParams count_occurrences_by_grade
46		#' @param refs (named `list` of `numeric`)\cr named list where each name corresponds to a reference grade level
47		#' and each entry represents a count.
48		#'
49		#' @return Formatted list of grade groupings.
50		#'
51		#' @examples
52		#' h_append_grade_groups(
53		#' list(
54		#' "Any Grade" = as.character(1:5),
55		#' "Grade 1-2" = c("1", "2"),
56		#' "Grade 3-4" = c("3", "4")
57		#' ),
58		#' list("1" = 10, "2" = 20, "3" = 30, "4" = 40, "5" = 50)
59		#' )
60		#'
61		#' h_append_grade_groups(
62		#' list(
63		#' "Any Grade" = as.character(5:1),
64		#' "Grade A" = "5",
65		#' "Grade B" = c("4", "3")
66		#' ),
67		#' list("1" = 10, "2" = 20, "3" = 30, "4" = 40, "5" = 50)
68		#' )
69		#'
70		#' h_append_grade_groups(
71		#' list(
72		#' "Any Grade" = as.character(1:5),
73		#' "Grade 1-2" = c("1", "2"),
74		#' "Grade 3-4" = c("3", "4")
75		#' ),
76		#' list("1" = 10, "2" = 5, "3" = 0)
77		#' )
78		#'
79		#' @export
80		h_append_grade_groups <- function(grade_groups, refs, remove_single = TRUE, only_grade_groups = FALSE) {
81	22x	checkmate::assert_list(grade_groups)
82	22x	checkmate::assert_list(refs)
83	22x	refs_orig <- refs
84	22x	elements <- unique(unlist(grade_groups))
85
86		### compute sums in groups
87	22x	grp_sum <- lapply(grade_groups, function(i) do.call(sum, refs[i]))
88	22x	if (!checkmate::test_subset(elements, names(refs))) {
89	2x	padding_el <- setdiff(elements, names(refs))
90	2x	refs[padding_el] <- 0
91		}
92	22x	result <- c(grp_sum, refs)
93
94		### order result while keeping grade_groups's ordering
95	22x	ordr <- grade_groups
96
97		# elements of any-grade group (if any) will be moved to the end
98	22x	is_any <- sapply(grade_groups, setequal, y = names(refs))
99	22x	ordr[is_any] <- list(character(0)) # hide elements under any-grade group
100
101		# groups-elements combined sequence
102	22x	ordr <- c(lapply(names(ordr), function(g) c(g, ordr[[g]])), recursive = TRUE, use.names = FALSE)
103	22x	ordr <- ordr[!duplicated(ordr)]
104
105		# append remaining elements (if any)
106	22x	ordr <- union(ordr, unlist(grade_groups[is_any])) # from any-grade group
107	22x	ordr <- union(ordr, names(refs)) # from refs
108
109		# remove elements of single-element groups, if any
110	22x	if (only_grade_groups) {
111	3x	ordr <- intersect(ordr, names(grade_groups))
112	19x	} else if (remove_single) {
113	19x	is_single <- sapply(grade_groups, length) == 1L
114	19x	ordr <- setdiff(ordr, unlist(grade_groups[is_single]))
115		}
116
117		# apply the order
118	22x	result <- result[ordr]
119
120		# remove groups without any elements in the original refs
121		# note: it's OK if groups have 0 value
122	22x	keep_grp <- vapply(grade_groups, function(x, rf) {
123	54x	any(x %in% rf)
124	22x	}, rf = names(refs_orig), logical(1))
125
126	22x	keep_el <- names(result) %in% names(refs_orig) \| names(result) %in% names(keep_grp)[keep_grp]
127	22x	result <- result[keep_el]
128
129	22x	result
130		}
131
132		#' @describeIn count_occurrences_by_grade Statistics function which counts the
133		#' number of patients by highest grade.
134		#'
135		#' @return
136		#' * `s_count_occurrences_by_grade()` returns a list of counts and fractions with one element per grade level or
137		#' grade level grouping.
138		#'
139		#' @examples
140		#' s_count_occurrences_by_grade(
141		#' df,
142		#' .N_col = 10L,
143		#' .var = "AETOXGR",
144		#' id = "USUBJID",
145		#' grade_groups = list("ANY" = levels(df$AETOXGR))
146		#' )
147		#'
148		#' @export
149		s_count_occurrences_by_grade <- function(df,
150		.var,
151		.N_col, # nolint
152		id = "USUBJID",
153		grade_groups = list(),
154		remove_single = TRUE,
155		only_grade_groups = FALSE,
156		labelstr = "") {
157	11x	assert_valid_factor(df[[.var]])
158	11x	assert_df_with_variables(df, list(grade = .var, id = id))
159
160	11x	if (nrow(df) < 1) {
161	2x	grade_levels <- levels(df[[.var]])
162	2x	l_count <- as.list(rep(0, length(grade_levels)))
163	2x	names(l_count) <- grade_levels
164		} else {
165	9x	if (isTRUE(is.factor(df[[id]]))) {
166	!	assert_valid_factor(df[[id]], any.missing = FALSE)
167		} else {
168	9x	checkmate::assert_character(df[[id]], min.chars = 1, any.missing = FALSE)
169		}
170	9x	checkmate::assert_count(.N_col)
171
172	9x	id <- df[[id]]
173	9x	grade <- df[[.var]]
174
175	9x	if (!is.ordered(grade)) {
176	9x	grade_lbl <- obj_label(grade)
177	9x	lvls <- levels(grade)
178	9x	if (sum(grepl("^\\d+$", lvls)) %in% c(0, length(lvls))) {
179	8x	lvl_ord <- lvls
180		} else {
181	1x	lvls[!grepl("^\\d+$", lvls)] <- min(as.numeric(lvls[grepl("^\\d+$", lvls)])) - 1
182	1x	lvl_ord <- levels(grade)[order(as.numeric(lvls))]
183		}
184	9x	grade <- formatters::with_label(factor(grade, levels = lvl_ord, ordered = TRUE), grade_lbl)
185		}
186
187	9x	missing_lvl <- grepl("missing", tolower(levels(grade)))
188	9x	if (any(missing_lvl)) {
189	1x	grade <- factor(
190	1x	grade,
191	1x	levels = c(levels(grade)[!missing_lvl], levels(grade)[missing_lvl]),
192	1x	ordered = is.ordered(grade)
193		)
194		}
195	9x	df_max <- stats::aggregate(grade ~ id, FUN = max, drop = FALSE)
196	9x	l_count <- as.list(table(df_max$grade))
197		}
198
199	11x	if (length(grade_groups) > 0) {
200	6x	l_count <- h_append_grade_groups(grade_groups, l_count, remove_single, only_grade_groups)
201		}
202
203	11x	l_count_fraction <- lapply(l_count, function(i, denom) c(i, i / denom), denom = .N_col)
204
205	11x	list(
206	11x	count_fraction = l_count_fraction
207		)
208		}
209
210		#' @describeIn count_occurrences_by_grade Formatted analysis function which is used as `afun`
211		#' in `count_occurrences_by_grade()`.
212		#'
213		#' @return
214		#' * `a_count_occurrences_by_grade()` returns the corresponding list with formatted [rtables::CellValue()].
215		#'
216		#' @examples
217		#' # We need to ungroup `count_fraction` first so that the `rtables` formatting
218		#' # function `format_count_fraction()` can be applied correctly.
219		#' afun <- make_afun(a_count_occurrences_by_grade, .ungroup_stats = "count_fraction")
220		#' afun(
221		#' df,
222		#' .N_col = 10L,
223		#' .var = "AETOXGR",
224		#' id = "USUBJID",
225		#' grade_groups = list("ANY" = levels(df$AETOXGR))
226		#' )
227		#'
228		#' @export
229		a_count_occurrences_by_grade <- make_afun(
230		s_count_occurrences_by_grade,
231		.formats = c("count_fraction" = format_count_fraction_fixed_dp)
232		)
233
234		#' @describeIn count_occurrences_by_grade Layout-creating function which can take statistics function
235		#' arguments and additional format arguments. This function is a wrapper for [rtables::analyze()].
236		#'
237		#' @return
238		#' * `count_occurrences_by_grade()` returns a layout object suitable for passing to further layouting functions,
239		#' or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
240		#' the statistics from `s_count_occurrences_by_grade()` to the table layout.
241		#'
242		#' @examples
243		#' library(dplyr)
244		#'
245		#' df <- data.frame(
246		#' USUBJID = as.character(c(1:6, 1)),
247		#' ARM = factor(c("A", "A", "A", "B", "B", "B", "A"), levels = c("A", "B")),
248		#' AETOXGR = factor(c(1, 2, 3, 4, 1, 2, 3), levels = c(1:5)),
249		#' AESEV = factor(
250		#' x = c("MILD", "MODERATE", "SEVERE", "MILD", "MILD", "MODERATE", "SEVERE"),
251		#' levels = c("MILD", "MODERATE", "SEVERE")
252		#' ),
253		#' stringsAsFactors = FALSE
254		#' )
255		#'
256		#' df_adsl <- df %>%
257		#' select(USUBJID, ARM) %>%
258		#' unique()
259		#'
260		#' # Layout creating function with custom format.
261		#' basic_table() %>%
262		#' split_cols_by("ARM") %>%
263		#' add_colcounts() %>%
264		#' count_occurrences_by_grade(
265		#' var = "AESEV",
266		#' .formats = c("count_fraction" = "xx.xx (xx.xx%)")
267		#' ) %>%
268		#' build_table(df, alt_counts_df = df_adsl)
269		#'
270		#' # Define additional grade groupings.
271		#' grade_groups <- list(
272		#' "-Any-" = c("1", "2", "3", "4", "5"),
273		#' "Grade 1-2" = c("1", "2"),
274		#' "Grade 3-5" = c("3", "4", "5")
275		#' )
276		#'
277		#' basic_table() %>%
278		#' split_cols_by("ARM") %>%
279		#' add_colcounts() %>%
280		#' count_occurrences_by_grade(
281		#' var = "AETOXGR",
282		#' grade_groups = grade_groups,
283		#' only_grade_groups = TRUE
284		#' ) %>%
285		#' build_table(df, alt_counts_df = df_adsl)
286		#'
287		#' @export
288		#' @order 2
289		count_occurrences_by_grade <- function(lyt,
290		var,
291		id = "USUBJID",
292		grade_groups = list(),
293		remove_single = TRUE,
294		only_grade_groups = FALSE,
295		var_labels = var,
296		show_labels = "default",
297		riskdiff = FALSE,
298		na_str = default_na_str(),
299		nested = TRUE,
300		...,
301		table_names = var,
302		.stats = NULL,
303		.formats = NULL,
304		.indent_mods = NULL,
305		.labels = NULL) {
306	11x	checkmate::assert_flag(riskdiff)
307	11x	s_args <- list(
308	11x	id = id, grade_groups = grade_groups, remove_single = remove_single, only_grade_groups = only_grade_groups, ...
309		)
310
311	11x	afun <- make_afun(
312	11x	a_count_occurrences_by_grade,
313	11x	.stats = .stats,
314	11x	.formats = .formats,
315	11x	.indent_mods = .indent_mods,
316	11x	.ungroup_stats = "count_fraction"
317		)
318
319	11x	extra_args <- if (isFALSE(riskdiff)) {
320	9x	s_args
321		} else {
322	2x	list(
323	2x	afun = list("s_count_occurrences_by_grade" = afun),
324	2x	.stats = .stats,
325	2x	.indent_mods = .indent_mods,
326	2x	s_args = s_args
327		)
328		}
329
330	11x	analyze(
331	11x	lyt = lyt,
332	11x	vars = var,
333	11x	var_labels = var_labels,
334	11x	show_labels = show_labels,
335	11x	afun = ifelse(isFALSE(riskdiff), afun, afun_riskdiff),
336	11x	table_names = table_names,
337	11x	na_str = na_str,
338	11x	nested = nested,
339	11x	extra_args = extra_args
340		)
341		}
342
343		#' @describeIn count_occurrences_by_grade Layout-creating function which can take content function arguments
344		#' and additional format arguments. This function is a wrapper for [rtables::summarize_row_groups()].
345		#'
346		#' @return
347		#' * `summarize_occurrences_by_grade()` returns a layout object suitable for passing to further layouting functions,
348		#' or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted content rows
349		#' containing the statistics from `s_count_occurrences_by_grade()` to the table layout.
350		#'
351		#' @examples
352		#' # Layout creating function with custom format.
353		#' basic_table() %>%
354		#' add_colcounts() %>%
355		#' split_rows_by("ARM", child_labels = "visible", nested = TRUE) %>%
356		#' summarize_occurrences_by_grade(
357		#' var = "AESEV",
358		#' .formats = c("count_fraction" = "xx.xx (xx.xx%)")
359		#' ) %>%
360		#' build_table(df, alt_counts_df = df_adsl)
361		#'
362		#' basic_table() %>%
363		#' add_colcounts() %>%
364		#' split_rows_by("ARM", child_labels = "visible", nested = TRUE) %>%
365		#' summarize_occurrences_by_grade(
366		#' var = "AETOXGR",
367		#' grade_groups = grade_groups
368		#' ) %>%
369		#' build_table(df, alt_counts_df = df_adsl)
370		#'
371		#' @export
372		#' @order 3
373		summarize_occurrences_by_grade <- function(lyt,
374		var,
375		id = "USUBJID",
376		grade_groups = list(),
377		remove_single = TRUE,
378		only_grade_groups = FALSE,
379		na_str = default_na_str(),
380		...,
381		.stats = NULL,
382		.formats = NULL,
383		.indent_mods = NULL,
384		.labels = NULL) {
385	3x	extra_args <- list(
386	3x	id = id, grade_groups = grade_groups, remove_single = remove_single, only_grade_groups = only_grade_groups, ...
387		)
388
389	3x	cfun <- make_afun(
390	3x	a_count_occurrences_by_grade,
391	3x	.stats = .stats,
392	3x	.formats = .formats,
393	3x	.labels = .labels,
394	3x	.indent_mods = .indent_mods,
395	3x	.ungroup_stats = "count_fraction"
396		)
397
398	3x	summarize_row_groups(
399	3x	lyt = lyt,
400	3x	var = var,
401	3x	cfun = cfun,
402	3x	na_str = na_str,
403	3x	extra_args = extra_args
404		)
405		}

1		#' Control function for subgroup treatment effect pattern (STEP) calculations
2		#'
3		#' @description `r lifecycle::badge("stable")`
4		#'
5		#' This is an auxiliary function for controlling arguments for STEP calculations.
6		#'
7		#' @param biomarker (`numeric` or `NULL`)\cr optional provision of the numeric biomarker variable, which
8		#' could be used to infer `bandwidth`, see below.
9		#' @param use_percentile (`flag`)\cr if `TRUE`, the running windows are created according to
10		#' quantiles rather than actual values, i.e. the bandwidth refers to the percentage of data
11		#' covered in each window. Suggest `TRUE` if the biomarker variable is not uniformly
12		#' distributed.
13		#' @param bandwidth (`numeric(1)` or `NULL`)\cr indicating the bandwidth of each window.
14		#' Depending on the argument `use_percentile`, it can be either the length of actual-value
15		#' windows on the real biomarker scale, or percentage windows.
16		#' If `use_percentile = TRUE`, it should be a number between 0 and 1.
17		#' If `NULL`, treat the bandwidth to be infinity, which means only one global model will be fitted.
18		#' By default, `0.25` is used for percentage windows and one quarter of the range of the `biomarker`
19		#' variable for actual-value windows.
20		#' @param degree (`integer(1)`)\cr the degree of polynomial function of the biomarker as an interaction term
21		#' with the treatment arm fitted at each window. If 0 (default), then the biomarker variable
22		#' is not included in the model fitted in each biomarker window.
23		#' @param num_points (`integer(1)`)\cr the number of points at which the hazard ratios are estimated. The
24		#' smallest number is 2.
25		#'
26		#' @return A list of components with the same names as the arguments, except `biomarker` which is
27		#' just used to calculate the `bandwidth` in case that actual biomarker windows are requested.
28		#'
29		#' @examples
30		#' # Provide biomarker values and request actual values to be used,
31		#' # so that bandwidth is chosen from range.
32		#' control_step(biomarker = 1:10, use_percentile = FALSE)
33		#'
34		#' # Use a global model with quadratic biomarker interaction term.
35		#' control_step(bandwidth = NULL, degree = 2)
36		#'
37		#' # Reduce number of points to be used.
38		#' control_step(num_points = 10)
39		#'
40		#' @export
41		control_step <- function(biomarker = NULL,
42		use_percentile = TRUE,
43		bandwidth,
44		degree = 0L,
45		num_points = 39L) {
46	31x	checkmate::assert_numeric(biomarker, null.ok = TRUE)
47	30x	checkmate::assert_flag(use_percentile)
48	30x	checkmate::assert_int(num_points, lower = 2)
49	29x	checkmate::assert_count(degree)
50
51	29x	if (missing(bandwidth)) {
52		# Infer bandwidth
53	21x	bandwidth <- if (use_percentile) {
54	18x	0.25
55	21x	} else if (!is.null(biomarker)) {
56	3x	diff(range(biomarker, na.rm = TRUE)) / 4
57		} else {
58	!	NULL
59		}
60		} else {
61		# Check bandwidth
62	8x	if (!is.null(bandwidth)) {
63	5x	if (use_percentile) {
64	4x	assert_proportion_value(bandwidth)
65		} else {
66	1x	checkmate::assert_scalar(bandwidth)
67	1x	checkmate::assert_true(bandwidth > 0)
68		}
69		}
70		}
71	28x	list(
72	28x	use_percentile = use_percentile,
73	28x	bandwidth = bandwidth,
74	28x	degree = as.integer(degree),
75	28x	num_points = as.integer(num_points)
76		)
77		}

1		#' Count number of patients
2		#'
3		#' @description `r lifecycle::badge("stable")`
4		#'
5		#' The analyze function [analyze_num_patients()] creates a layout element to count total numbers of unique or
6		#' non-unique patients. The primary analysis variable `vars` is used to uniquely identify patients.
7		#'
8		#' The `count_by` variable can be used to identify non-unique patients such that the number of patients with a unique
9		#' combination of values in `vars` and `count_by` will be returned instead as the `nonunique` statistic. The `required`
10		#' variable can be used to specify a variable required to be non-missing for the record to be included in the counts.
11		#'
12		#' The summarize function [summarize_num_patients()] performs the same function as [analyze_num_patients()] except it
13		#' creates content rows, not data rows, to summarize the current table row/column context and operates on the level of
14		#' the latest row split or the root of the table if no row splits have occurred.
15		#'
16		#' @inheritParams argument_convention
17		#' @param required (`character` or `NULL`)\cr name of a variable that is required to be non-missing.
18		#' @param count_by (`character` or `NULL`)\cr name of a variable to be combined with `vars` when counting
19		#' `nonunique` records.
20		#' @param unique_count_suffix (`flag`)\cr whether the `"(n)"` suffix should be added to `unique_count` labels.
21		#' Defaults to `TRUE`.
22		#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("summarize_num_patients")`
23		#' to see available statistics for this function.
24		#'
25		#' @name summarize_num_patients
26		#' @order 1
27		NULL
28
29		#' @describeIn summarize_num_patients Statistics function which counts the number of
30		#' unique patients, the corresponding percentage taken with respect to the
31		#' total number of patients, and the number of non-unique patients.
32		#'
33		#' @param x (`character` or `factor`)\cr vector of patient IDs.
34		#'
35		#' @return
36		#' * `s_num_patients()` returns a named `list` of 3 statistics:
37		#' * `unique`: Vector of counts and percentages.
38		#' * `nonunique`: Vector of counts.
39		#' * `unique_count`: Counts.
40		#'
41		#' @examples
42		#' # Use the statistics function to count number of unique and nonunique patients.
43		#' s_num_patients(x = as.character(c(1, 1, 1, 2, 4, NA)), labelstr = "", .N_col = 6L)
44		#' s_num_patients(
45		#' x = as.character(c(1, 1, 1, 2, 4, NA)),
46		#' labelstr = "",
47		#' .N_col = 6L,
48		#' count_by = c(1, 1, 2, 1, 1, 1)
49		#' )
50		#'
51		#' @export
52		s_num_patients <- function(x, labelstr, .N_col, count_by = NULL, unique_count_suffix = TRUE) { # nolint
53
54	146x	checkmate::assert_string(labelstr)
55	146x	checkmate::assert_count(.N_col)
56	146x	checkmate::assert_multi_class(x, classes = c("factor", "character"))
57	146x	checkmate::assert_flag(unique_count_suffix)
58
59	146x	count1 <- n_available(unique(x))
60	146x	count2 <- n_available(x)
61
62	146x	if (!is.null(count_by)) {
63	16x	checkmate::assert_vector(count_by, len = length(x))
64	16x	count2 <- n_available(unique(interaction(x, count_by)))
65		}
66
67	146x	out <- list(
68	146x	unique = formatters::with_label(c(count1, ifelse(count1 == 0 && .N_col == 0, 0, count1 / .N_col)), labelstr),
69	146x	nonunique = formatters::with_label(count2, labelstr),
70	146x	unique_count = formatters::with_label(
71	146x	count1, ifelse(unique_count_suffix, paste0(labelstr, if (nzchar(labelstr)) " ", "(n)"), labelstr)
72		)
73		)
74
75	146x	out
76		}
77
78		#' @describeIn summarize_num_patients Statistics function which counts the number of unique patients
79		#' in a column (variable), the corresponding percentage taken with respect to the total number of
80		#' patients, and the number of non-unique patients in the column.
81		#'
82		#' @return
83		#' * `s_num_patients_content()` returns the same values as `s_num_patients()`.
84		#'
85		#' @examples
86		#' # Count number of unique and non-unique patients.
87		#'
88		#' df <- data.frame(
89		#' USUBJID = as.character(c(1, 2, 1, 4, NA)),
90		#' EVENT = as.character(c(10, 15, 10, 17, 8))
91		#' )
92		#' s_num_patients_content(df, .N_col = 5, .var = "USUBJID")
93		#'
94		#' df_by_event <- data.frame(
95		#' USUBJID = as.character(c(1, 2, 1, 4, NA)),
96		#' EVENT = c(10, 15, 10, 17, 8)
97		#' )
98		#' s_num_patients_content(df_by_event, .N_col = 5, .var = "USUBJID", count_by = "EVENT")
99		#'
100		#' @export
101		s_num_patients_content <- function(df,
102		labelstr = "",
103		.N_col, # nolint
104		.var,
105		required = NULL,
106		count_by = NULL,
107		unique_count_suffix = TRUE) {
108	56x	checkmate::assert_string(.var)
109	56x	checkmate::assert_data_frame(df)
110	56x	if (is.null(count_by)) {
111	53x	assert_df_with_variables(df, list(id = .var))
112		} else {
113	3x	assert_df_with_variables(df, list(id = .var, count_by = count_by))
114		}
115	56x	if (!is.null(required)) {
116	!	checkmate::assert_string(required)
117	!	assert_df_with_variables(df, list(required = required))
118	!	df <- df[!is.na(df[[required]]), , drop = FALSE]
119		}
120
121	56x	x <- df[[.var]]
122	56x	y <- if (is.null(count_by)) NULL else df[[count_by]]
123
124	56x	s_num_patients(
125	56x	x = x,
126	56x	labelstr = labelstr,
127	56x	.N_col = .N_col,
128	56x	count_by = y,
129	56x	unique_count_suffix = unique_count_suffix
130		)
131		}
132
133		c_num_patients <- make_afun(
134		s_num_patients_content,
135		.stats = c("unique", "nonunique", "unique_count"),
136		.formats = c(unique = format_count_fraction_fixed_dp, nonunique = "xx", unique_count = "xx")
137		)
138
139		#' @describeIn summarize_num_patients Layout-creating function which can take statistics function arguments
140		#' and additional format arguments. This function is a wrapper for [rtables::summarize_row_groups()].
141		#'
142		#' @return
143		#' * `summarize_num_patients()` returns a layout object suitable for passing to further layouting functions,
144		#' or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
145		#' the statistics from `s_num_patients_content()` to the table layout.
146		#'
147		#' @examples
148		#' # summarize_num_patients
149		#' tbl <- basic_table() %>%
150		#' split_cols_by("ARM") %>%
151		#' split_rows_by("SEX") %>%
152		#' summarize_num_patients("USUBJID", .stats = "unique_count") %>%
153		#' build_table(df)
154		#'
155		#' tbl
156		#'
157		#' @export
158		#' @order 3
159		summarize_num_patients <- function(lyt,
160		var,
161		required = NULL,
162		count_by = NULL,
163		unique_count_suffix = TRUE,
164		na_str = default_na_str(),
165		.stats = NULL,
166		.formats = NULL,
167		.labels = c(
168		unique = "Number of patients with at least one event",
169		nonunique = "Number of events"
170		),
171		.indent_mods = 0L,
172		riskdiff = FALSE,
173		...) {
174	16x	checkmate::assert_flag(riskdiff)
175
176	5x	if (is.null(.stats)) .stats <- c("unique", "nonunique", "unique_count")
177	8x	if (length(.labels) > length(.stats)) .labels <- .labels[names(.labels) %in% .stats]
178
179	16x	s_args <- list(required = required, count_by = count_by, unique_count_suffix = unique_count_suffix, ...)
180
181	16x	cfun <- make_afun(
182	16x	c_num_patients,
183	16x	.stats = .stats,
184	16x	.formats = .formats,
185	16x	.labels = .labels
186		)
187
188	16x	extra_args <- if (isFALSE(riskdiff)) {
189	14x	s_args
190		} else {
191	2x	list(
192	2x	afun = list("s_num_patients_content" = cfun),
193	2x	.stats = .stats,
194	2x	.indent_mods = .indent_mods,
195	2x	s_args = s_args
196		)
197		}
198
199	16x	summarize_row_groups(
200	16x	lyt = lyt,
201	16x	var = var,
202	16x	cfun = ifelse(isFALSE(riskdiff), cfun, afun_riskdiff),
203	16x	na_str = na_str,
204	16x	extra_args = extra_args,
205	16x	indent_mod = .indent_mods
206		)
207		}
208
209		#' @describeIn summarize_num_patients Layout-creating function which can take statistics function arguments
210		#' and additional format arguments. This function is a wrapper for [rtables::analyze()].
211		#'
212		#' @return
213		#' * `analyze_num_patients()` returns a layout object suitable for passing to further layouting functions,
214		#' or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
215		#' the statistics from `s_num_patients_content()` to the table layout.
216		#'
217		#' @details In general, functions that starts with `analyze*` are expected to
218		#' work like [rtables::analyze()], while functions that starts with `summarize*`
219		#' are based upon [rtables::summarize_row_groups()]. The latter provides a
220		#' value for each dividing split in the row and column space, but, being it
221		#' bound to the fundamental splits, it is repeated by design in every page
222		#' when pagination is involved.
223		#'
224		#' @note As opposed to [summarize_num_patients()], this function does not repeat the produced rows.
225		#'
226		#' @examples
227		#' df <- data.frame(
228		#' USUBJID = as.character(c(1, 2, 1, 4, NA, 6, 6, 8, 9)),
229		#' ARM = c("A", "A", "A", "A", "A", "B", "B", "B", "B"),
230		#' AGE = c(10, 15, 10, 17, 8, 11, 11, 19, 17),
231		#' SEX = c("M", "M", "M", "F", "F", "F", "M", "F", "M")
232		#' )
233		#'
234		#' # analyze_num_patients
235		#' tbl <- basic_table() %>%
236		#' split_cols_by("ARM") %>%
237		#' add_colcounts() %>%
238		#' analyze_num_patients("USUBJID", .stats = c("unique")) %>%
239		#' build_table(df)
240		#'
241		#' tbl
242		#'
243		#' @export
244		#' @order 2
245		analyze_num_patients <- function(lyt,
246		vars,
247		required = NULL,
248		count_by = NULL,
249		unique_count_suffix = TRUE,
250		na_str = default_na_str(),
251		nested = TRUE,
252		.stats = NULL,
253		.formats = NULL,
254		.labels = c(
255		unique = "Number of patients with at least one event",
256		nonunique = "Number of events"
257		),
258		show_labels = c("default", "visible", "hidden"),
259		.indent_mods = 0L,
260		riskdiff = FALSE,
261		...) {
262	4x	checkmate::assert_flag(riskdiff)
263
264	1x	if (is.null(.stats)) .stats <- c("unique", "nonunique", "unique_count")
265	!	if (length(.labels) > length(.stats)) .labels <- .labels[names(.labels) %in% .stats]
266
267	4x	s_args <- list(required = required, count_by = count_by, unique_count_suffix = unique_count_suffix, ...)
268
269	4x	afun <- make_afun(
270	4x	c_num_patients,
271	4x	.stats = .stats,
272	4x	.formats = .formats,
273	4x	.labels = .labels
274		)
275
276	4x	extra_args <- if (isFALSE(riskdiff)) {
277	2x	s_args
278		} else {
279	2x	list(
280	2x	afun = list("s_num_patients_content" = afun),
281	2x	.stats = .stats,
282	2x	.indent_mods = .indent_mods,
283	2x	s_args = s_args
284		)
285		}
286
287	4x	analyze(
288	4x	afun = ifelse(isFALSE(riskdiff), afun, afun_riskdiff),
289	4x	lyt = lyt,
290	4x	vars = vars,
291	4x	na_str = na_str,
292	4x	nested = nested,
293	4x	extra_args = extra_args,
294	4x	show_labels = show_labels,
295	4x	indent_mod = .indent_mods
296		)
297		}

1		#' Re-implemented `range()` default S3 method for numerical objects
2		#'
3		#' This function returns `c(NA, NA)` instead of `c(-Inf, Inf)` for zero-length data
4		#' without any warnings.
5		#'
6		#' @param x (`numeric`)\cr a sequence of numbers for which the range is computed.
7		#' @param na.rm (`flag`)\cr flag indicating if `NA` should be omitted.
8		#' @param finite (`flag`)\cr flag indicating if non-finite elements should be removed.
9		#'
10		#' @return A 2-element vector of class `numeric`.
11		#'
12		#' @keywords internal
13		range_noinf <- function(x, na.rm = FALSE, finite = FALSE) { # nolint
14
15	1842x	checkmate::assert_numeric(x)
16
17	1842x	if (finite) {
18	24x	x <- x[is.finite(x)] # removes NAs too
19	1818x	} else if (na.rm) {
20	708x	x <- x[!is.na(x)]
21		}
22
23	1842x	if (length(x) == 0) {
24	111x	rval <- c(NA, NA)
25	111x	mode(rval) <- typeof(x)
26		} else {
27	1731x	rval <- c(min(x, na.rm = FALSE), max(x, na.rm = FALSE))
28		}
29
30	1842x	return(rval)
31		}
32
33		#' Utility function to create label for confidence interval
34		#'
35		#' @description `r lifecycle::badge("stable")`
36		#'
37		#' @inheritParams argument_convention
38		#'
39		#' @return A `string`.
40		#'
41		#' @export
42		f_conf_level <- function(conf_level) {
43	3860x	assert_proportion_value(conf_level)
44	3858x	paste0(conf_level * 100, "% CI")
45		}
46
47		#' Utility function to create label for p-value
48		#'
49		#' @description `r lifecycle::badge("stable")`
50		#'
51		#' @param test_mean (`numeric(1)`)\cr mean value to test under the null hypothesis.
52		#'
53		#' @return A `string`.
54		#'
55		#' @export
56		f_pval <- function(test_mean) {
57	1103x	checkmate::assert_numeric(test_mean, len = 1)
58	1101x	paste0("p-value (H0: mean = ", test_mean, ")")
59		}
60
61		#' Utility function to return a named list of covariate names
62		#'
63		#' @param covariates (`character`)\cr a vector that can contain single variable names (such as
64		#' `"X1"`), and/or interaction terms indicated by `"X1 * X2"`.
65		#'
66		#' @return A named `list` of `character` vector.
67		#'
68		#' @keywords internal
69		get_covariates <- function(covariates) {
70	14x	checkmate::assert_character(covariates)
71	12x	cov_vars <- unique(trimws(unlist(strsplit(covariates, "\\*"))))
72	12x	stats::setNames(as.list(cov_vars), cov_vars)
73		}
74
75		#' Replicate entries of a vector if required
76		#'
77		#' @description `r lifecycle::badge("stable")`
78		#'
79		#' Replicate entries of a vector if required.
80		#'
81		#' @inheritParams argument_convention
82		#' @param n (`integer(1)`)\cr number of entries that are needed.
83		#'
84		#' @return `x` if it has the required length already or is `NULL`,
85		#' otherwise if it is scalar the replicated version of it with `n` entries.
86		#'
87		#' @note This function will fail if `x` is not of length `n` and/or is not a scalar.
88		#'
89		#' @export
90		to_n <- function(x, n) {
91	5x	if (is.null(x)) {
92	1x	NULL
93	4x	} else if (length(x) == 1) {
94	1x	rep(x, n)
95	3x	} else if (length(x) == n) {
96	2x	x
97		} else {
98	1x	stop("dimension mismatch")
99		}
100		}
101
102		#' Check element dimension
103		#'
104		#' Checks if the elements in `...` have the same dimension.
105		#'
106		#' @param ... (`data.frame` or `vector`)\cr any data frames or vectors.
107		#' @param omit_null (`flag`)\cr whether `NULL` elements in `...` should be omitted from the check.
108		#'
109		#' @return A `logical` value.
110		#'
111		#' @keywords internal
112		check_same_n <- function(..., omit_null = TRUE) {
113	2x	dots <- list(...)
114
115	2x	n_list <- Map(
116	2x	function(x, name) {
117	5x	if (is.null(x)) {
118	!	if (omit_null) {
119	2x	NA_integer_
120		} else {
121	!	stop("arg", name, "is not supposed to be NULL")
122		}
123	5x	} else if (is.data.frame(x)) {
124	!	nrow(x)
125	5x	} else if (is.atomic(x)) {
126	5x	length(x)
127		} else {
128	!	stop("data structure for ", name, "is currently not supported")
129		}
130		},
131	2x	dots, names(dots)
132		)
133
134	2x	n <- stats::na.omit(unlist(n_list))
135
136	2x	if (length(unique(n)) > 1) {
137	!	sel <- which(n != n[1])
138	!	stop("Dimension mismatch:", paste(names(n)[sel], collapse = ", "), " do not have N=", n[1])
139		}
140
141	2x	TRUE
142		}
143
144		#' Utility function to check if a float value is equal to another float value
145		#'
146		#' Uses `.Machine$double.eps` as the tolerance for the comparison.
147		#'
148		#' @param x (`numeric(1)`)\cr a float number.
149		#' @param y (`numeric(1)`)\cr a float number.
150		#'
151		#' @return `TRUE` if identical, otherwise `FALSE`.
152		#'
153		#' @keywords internal
154		.is_equal_float <- function(x, y) {
155	2820x	checkmate::assert_number(x)
156	2820x	checkmate::assert_number(y)
157
158		# Define a tolerance
159	2820x	tolerance <- .Machine$double.eps
160
161		# Check if x is close enough to y
162	2820x	abs(x - y) < tolerance
163		}
164
165		#' Make names without dots
166		#'
167		#' @param nams (`character`)\cr vector of original names.
168		#'
169		#' @return A `character` `vector` of proper names, which does not use dots in contrast to [make.names()].
170		#'
171		#' @keywords internal
172		make_names <- function(nams) {
173	6x	orig <- make.names(nams)
174	6x	gsub(".", "", x = orig, fixed = TRUE)
175		}
176
177		#' Conversion of months to days
178		#'
179		#' @description `r lifecycle::badge("stable")`
180		#'
181		#' Conversion of months to days. This is an approximative calculation because it
182		#' considers each month as having an average of 30.4375 days.
183		#'
184		#' @param x (`numeric(1)`)\cr time in months.
185		#'
186		#' @return A `numeric` vector with the time in days.
187		#'
188		#' @examples
189		#' x <- c(13.25, 8.15, 1, 2.834)
190		#' month2day(x)
191		#'
192		#' @export
193		month2day <- function(x) {
194	1x	checkmate::assert_numeric(x)
195	1x	x * 30.4375
196		}
197
198		#' Conversion of days to months
199		#'
200		#' @param x (`numeric(1)`)\cr time in days.
201		#'
202		#' @return A `numeric` vector with the time in months.
203		#'
204		#' @examples
205		#' x <- c(403, 248, 30, 86)
206		#' day2month(x)
207		#'
208		#' @export
209		day2month <- function(x) {
210	19x	checkmate::assert_numeric(x)
211	19x	x / 30.4375
212		}
213
214		#' Return an empty numeric if all elements are `NA`.
215		#'
216		#' @param x (`numeric`)\cr vector.
217		#'
218		#' @return An empty `numeric` if all elements of `x` are `NA`, otherwise `x`.
219		#'
220		#' @examples
221		#' x <- c(NA, NA, NA)
222		#' # Internal function - empty_vector_if_na
223		#' @keywords internal
224		empty_vector_if_na <- function(x) {
225	1017x	if (all(is.na(x))) {
226	310x	numeric()
227		} else {
228	707x	x
229		}
230		}
231
232		#' Element-wise combination of two vectors
233		#'
234		#' @param x (`vector`)\cr first vector to combine.
235		#' @param y (`vector`)\cr second vector to combine.
236		#'
237		#' @return A `list` where each element combines corresponding elements of `x` and `y`.
238		#'
239		#' @examples
240		#' combine_vectors(1:3, 4:6)
241		#'
242		#' @export
243		combine_vectors <- function(x, y) {
244	51x	checkmate::assert_vector(x)
245	51x	checkmate::assert_vector(y, len = length(x))
246
247	51x	result <- lapply(as.data.frame(rbind(x, y)), `c`)
248	51x	names(result) <- NULL
249	51x	result
250		}
251
252		#' Extract elements by name
253		#'
254		#' This utility function extracts elements from a vector `x` by `names`.
255		#' Differences to the standard `[` function are:
256		#'
257		#' - If `x` is `NULL`, then still always `NULL` is returned (same as in base function).
258		#' - If `x` is not `NULL`, then the intersection of its names is made with `names` and those
259		#' elements are returned. That is, `names` which don't appear in `x` are not returned as `NA`s.
260		#'
261		#' @param x (named `vector`)\cr where to extract named elements from.
262		#' @param names (`character`)\cr vector of names to extract.
263		#'
264		#' @return `NULL` if `x` is `NULL`, otherwise the extracted elements from `x`.
265		#'
266		#' @keywords internal
267		extract_by_name <- function(x, names) {
268	3x	if (is.null(x)) {
269	1x	return(NULL)
270		}
271	2x	checkmate::assert_named(x)
272	2x	checkmate::assert_character(names)
273	2x	which_extract <- intersect(names(x), names)
274	2x	if (length(which_extract) > 0) {
275	1x	x[which_extract]
276		} else {
277	1x	NULL
278		}
279		}
280
281		#' Labels for adverse event baskets
282		#'
283		#' @description `r lifecycle::badge("stable")`
284		#'
285		#' @param aesi (`character`)\cr vector with standardized MedDRA query name (e.g. `SMQxxNAM`) or customized query
286		#' name (e.g. `CQxxNAM`).
287		#' @param scope (`character`)\cr vector with scope of query (e.g. `SMQxxSC`).
288		#'
289		#' @return A `string` with the standard label for the AE basket.
290		#'
291		#' @examples
292		#' adae <- tern_ex_adae
293		#'
294		#' # Standardized query label includes scope.
295		#' aesi_label(adae$SMQ01NAM, scope = adae$SMQ01SC)
296		#'
297		#' # Customized query label.
298		#' aesi_label(adae$CQ01NAM)
299		#'
300		#' @export
301		aesi_label <- function(aesi, scope = NULL) {
302	4x	checkmate::assert_character(aesi)
303	4x	checkmate::assert_character(scope, null.ok = TRUE)
304	4x	aesi_label <- obj_label(aesi)
305	4x	aesi <- sas_na(aesi)
306	4x	aesi <- unique(aesi)[!is.na(unique(aesi))]
307
308	4x	lbl <- if (length(aesi) == 1 && !is.null(scope)) {
309	1x	scope <- sas_na(scope)
310	1x	scope <- unique(scope)[!is.na(unique(scope))]
311	1x	checkmate::assert_string(scope)
312	1x	paste0(aesi, " (", scope, ")")
313	4x	} else if (length(aesi) == 1 && is.null(scope)) {
314	1x	aesi
315		} else {
316	2x	aesi_label
317		}
318
319	4x	lbl
320		}
321
322		#' Indicate study arm variable in formula
323		#'
324		#' We use `study_arm` to indicate the study arm variable in `tern` formulas.
325		#'
326		#' @param x arm information
327		#'
328		#' @return `x`
329		#'
330		#' @keywords internal
331		study_arm <- function(x) {
332	!	structure(x, varname = deparse(substitute(x)))
333		}
334
335		#' Smooth function with optional grouping
336		#'
337		#' @description `r lifecycle::badge("stable")`
338		#'
339		#' This produces `loess` smoothed estimates of `y` with Student confidence intervals.
340		#'
341		#' @param df (`data.frame`)\cr data set containing all analysis variables.
342		#' @param x (`string`)\cr x column name.
343		#' @param y (`string`)\cr y column name.
344		#' @param groups (`character` or `NULL`)\cr vector with optional grouping variables names.
345		#' @param level (`proportion`)\cr level of confidence interval to use (0.95 by default).
346		#'
347		#' @return A `data.frame` with original `x`, smoothed `y`, `ylow`, and `yhigh`, and
348		#' optional `groups` variables formatted as `factor` type.
349		#'
350		#' @export
351		get_smooths <- function(df, x, y, groups = NULL, level = 0.95) {
352	5x	checkmate::assert_data_frame(df)
353	5x	df_cols <- colnames(df)
354	5x	checkmate::assert_string(x)
355	5x	checkmate::assert_subset(x, df_cols)
356	5x	checkmate::assert_numeric(df[[x]])
357	5x	checkmate::assert_string(y)
358	5x	checkmate::assert_subset(y, df_cols)
359	5x	checkmate::assert_numeric(df[[y]])
360
361	5x	if (!is.null(groups)) {
362	4x	checkmate::assert_character(groups)
363	4x	checkmate::assert_subset(groups, df_cols)
364		}
365
366	5x	smooths <- function(x, y) {
367	18x	stats::predict(stats::loess(y ~ x), se = TRUE)
368		}
369
370	5x	if (!is.null(groups)) {
371	4x	cc <- stats::complete.cases(df[c(x, y, groups)])
372	4x	df_c <- df[cc, c(x, y, groups)]
373	4x	df_c_ordered <- df_c[do.call("order", as.list(df_c[, groups, drop = FALSE])), , drop = FALSE]
374	4x	df_c_g <- data.frame(Map(as.factor, df_c_ordered[groups]))
375
376	4x	df_smooth_raw <-
377	4x	by(df_c_ordered, df_c_g, function(d) {
378	17x	plx <- smooths(d[[x]], d[[y]])
379	17x	data.frame(
380	17x	x = d[[x]],
381	17x	y = plx$fit,
382	17x	ylow = plx$fit - stats::qt(level, plx$df) * plx$se.fit,
383	17x	yhigh = plx$fit + stats::qt(level, plx$df) * plx$se.fit
384		)
385		})
386
387	4x	df_smooth <- do.call(rbind, df_smooth_raw)
388	4x	df_smooth[groups] <- df_c_g
389
390	4x	df_smooth
391		} else {
392	1x	cc <- stats::complete.cases(df[c(x, y)])
393	1x	df_c <- df[cc, ]
394	1x	plx <- smooths(df_c[[x]], df_c[[y]])
395
396	1x	df_smooth <- data.frame(
397	1x	x = df_c[[x]],
398	1x	y = plx$fit,
399	1x	ylow = plx$fit - stats::qt(level, plx$df) * plx$se.fit,
400	1x	yhigh = plx$fit + stats::qt(level, plx$df) * plx$se.fit
401		)
402
403	1x	df_smooth
404		}
405		}
406
407		#' Number of available (non-missing entries) in a vector
408		#'
409		#' Small utility function for better readability.
410		#'
411		#' @param x (`vector`)\cr vector in which to count non-missing values.
412		#'
413		#' @return Number of non-missing values.
414		#'
415		#' @keywords internal
416		n_available <- function(x) {
417	351x	sum(!is.na(x))
418		}
419
420		#' Reapply variable labels
421		#'
422		#' This is a helper function that is used in tests.
423		#'
424		#' @param x (`vector`)\cr vector of elements that needs new labels.
425		#' @param varlabels (`character`)\cr vector of labels for `x`.
426		#' @param ... further parameters to be added to the list.
427		#'
428		#' @return `x` with variable labels reapplied.
429		#'
430		#' @export
431		reapply_varlabels <- function(x, varlabels, ...) {
432	11x	named_labels <- c(as.list(varlabels), list(...))
433	11x	formatters::var_labels(x)[names(named_labels)] <- as.character(named_labels)
434	11x	x
435		}
436
437		# Wrapper function of survival::clogit so that when model fitting failed, a more useful message would show
438		clogit_with_tryCatch <- function(formula, data, ...) { # nolint
439	30x	tryCatch(
440	30x	survival::clogit(formula = formula, data = data, ...),
441	30x	error = function(e) stop("model not built successfully with survival::clogit")
442		)
443		}

1		#' Confidence interval for mean
2		#'
3		#' @description `r lifecycle::badge("stable")`
4		#'
5		#' Convenient function for calculating the mean confidence interval. It calculates the arithmetic as well as the
6		#' geometric mean. It can be used as a `ggplot` helper function for plotting.
7		#'
8		#' @inheritParams argument_convention
9		#' @param n_min (`numeric(1)`)\cr a minimum number of non-missing `x` to estimate the confidence interval for mean.
10		#' @param gg_helper (`flag`)\cr whether output should be aligned for use with `ggplot`s.
11		#' @param geom_mean (`flag`)\cr whether the geometric mean should be calculated.
12		#'
13		#' @return A named `vector` of values `mean_ci_lwr` and `mean_ci_upr`.
14		#'
15		#' @examples
16		#' stat_mean_ci(sample(10), gg_helper = FALSE)
17		#'
18		#' p <- ggplot2::ggplot(mtcars, ggplot2::aes(cyl, mpg)) +
19		#' ggplot2::geom_point()
20		#'
21		#' p + ggplot2::stat_summary(
22		#' fun.data = stat_mean_ci,
23		#' geom = "errorbar"
24		#' )
25		#'
26		#' p + ggplot2::stat_summary(
27		#' fun.data = stat_mean_ci,
28		#' fun.args = list(conf_level = 0.5),
29		#' geom = "errorbar"
30		#' )
31		#'
32		#' p + ggplot2::stat_summary(
33		#' fun.data = stat_mean_ci,
34		#' fun.args = list(conf_level = 0.5, geom_mean = TRUE),
35		#' geom = "errorbar"
36		#' )
37		#'
38		#' @export
39		stat_mean_ci <- function(x,
40		conf_level = 0.95,
41		na.rm = TRUE, # nolint
42		n_min = 2,
43		gg_helper = TRUE,
44		geom_mean = FALSE) {
45	2211x	if (na.rm) {
46	10x	x <- stats::na.omit(x)
47		}
48	2211x	n <- length(x)
49
50	2211x	if (!geom_mean) {
51	1113x	m <- mean(x)
52		} else {
53	1098x	negative_values_exist <- any(is.na(x[!is.na(x)]) <- x[!is.na(x)] <= 0)
54	1098x	if (negative_values_exist) {
55	22x	m <- NA_real_
56		} else {
57	1076x	x <- log(x)
58	1076x	m <- mean(x)
59		}
60		}
61
62	2211x	if (n < n_min \|\| is.na(m)) {
63	302x	ci <- c(mean_ci_lwr = NA_real_, mean_ci_upr = NA_real_)
64		} else {
65	1909x	hci <- stats::qt((1 + conf_level) / 2, df = n - 1) * stats::sd(x) / sqrt(n)
66	1909x	ci <- c(mean_ci_lwr = m - hci, mean_ci_upr = m + hci)
67	1909x	if (geom_mean) {
68	945x	ci <- exp(ci)
69		}
70		}
71
72	2211x	if (gg_helper) {
73	4x	m <- ifelse(is.na(m), NA_real_, m)
74	4x	ci <- data.frame(y = ifelse(geom_mean, exp(m), m), ymin = ci[[1]], ymax = ci[[2]])
75		}
76
77	2211x	return(ci)
78		}
79
80		#' Confidence interval for median
81		#'
82		#' @description `r lifecycle::badge("stable")`
83		#'
84		#' Convenient function for calculating the median confidence interval. It can be used as a `ggplot` helper
85		#' function for plotting.
86		#'
87		#' @inheritParams argument_convention
88		#' @param gg_helper (`flag`)\cr whether output should be aligned for use with `ggplot`s.
89		#'
90		#' @details This function was adapted from `DescTools/versions/0.99.35/source`
91		#'
92		#' @return A named `vector` of values `median_ci_lwr` and `median_ci_upr`.
93		#'
94		#' @examples
95		#' stat_median_ci(sample(10), gg_helper = FALSE)
96		#'
97		#' p <- ggplot2::ggplot(mtcars, ggplot2::aes(cyl, mpg)) +
98		#' ggplot2::geom_point()
99		#' p + ggplot2::stat_summary(
100		#' fun.data = stat_median_ci,
101		#' geom = "errorbar"
102		#' )
103		#'
104		#' @export
105		stat_median_ci <- function(x,
106		conf_level = 0.95,
107		na.rm = TRUE, # nolint
108		gg_helper = TRUE) {
109	1111x	x <- unname(x)
110	1111x	if (na.rm) {
111	9x	x <- x[!is.na(x)]
112		}
113	1111x	n <- length(x)
114	1111x	med <- stats::median(x)
115
116	1111x	k <- stats::qbinom(p = (1 - conf_level) / 2, size = n, prob = 0.5, lower.tail = TRUE)
117
118		# k == 0 - for small samples (e.g. n <= 5) ci can be outside the observed range
119	1111x	if (k == 0 \|\| is.na(med)) {
120	242x	ci <- c(median_ci_lwr = NA_real_, median_ci_upr = NA_real_)
121	242x	empir_conf_level <- NA_real_
122		} else {
123	869x	x_sort <- sort(x)
124	869x	ci <- c(median_ci_lwr = x_sort[k], median_ci_upr = x_sort[n - k + 1])
125	869x	empir_conf_level <- 1 - 2 * stats::pbinom(k - 1, size = n, prob = 0.5)
126		}
127
128	1111x	if (gg_helper) {
129	4x	ci <- data.frame(y = med, ymin = ci[[1]], ymax = ci[[2]])
130		}
131
132	1111x	attr(ci, "conf_level") <- empir_conf_level
133
134	1111x	return(ci)
135		}
136
137		#' p-Value of the mean
138		#'
139		#' @description `r lifecycle::badge("stable")`
140		#'
141		#' Convenient function for calculating the two-sided p-value of the mean.
142		#'
143		#' @inheritParams argument_convention
144		#' @param n_min (`numeric(1)`)\cr a minimum number of non-missing `x` to estimate the p-value of the mean.
145		#' @param test_mean (`numeric(1)`)\cr mean value to test under the null hypothesis.
146		#'
147		#' @return A p-value.
148		#'
149		#' @examples
150		#' stat_mean_pval(sample(10))
151		#'
152		#' stat_mean_pval(rnorm(10), test_mean = 0.5)
153		#'
154		#' @export
155		stat_mean_pval <- function(x,
156		na.rm = TRUE, # nolint
157		n_min = 2,
158		test_mean = 0) {
159	1111x	if (na.rm) {
160	9x	x <- stats::na.omit(x)
161		}
162	1111x	n <- length(x)
163
164	1111x	x_mean <- mean(x)
165	1111x	x_sd <- stats::sd(x)
166
167	1111x	if (n < n_min) {
168	140x	pv <- c(p_value = NA_real_)
169		} else {
170	971x	x_se <- stats::sd(x) / sqrt(n)
171	971x	ttest <- (x_mean - test_mean) / x_se
172	971x	pv <- c(p_value = 2 * stats::pt(-abs(ttest), df = n - 1))
173		}
174
175	1111x	return(pv)
176		}
177
178		#' Proportion difference and confidence interval
179		#'
180		#' @description `r lifecycle::badge("stable")`
181		#'
182		#' Function for calculating the proportion (or risk) difference and confidence interval between arm
183		#' X (reference group) and arm Y. Risk difference is calculated by subtracting cumulative incidence
184		#' in arm Y from cumulative incidence in arm X.
185		#'
186		#' @inheritParams argument_convention
187		#' @param x (`list` of `integer`)\cr list of number of occurrences in arm X (reference group).
188		#' @param y (`list` of `integer`)\cr list of number of occurrences in arm Y. Must be of equal length to `x`.
189		#' @param N_x (`numeric(1)`)\cr total number of records in arm X.
190		#' @param N_y (`numeric(1)`)\cr total number of records in arm Y.
191		#' @param list_names (`character`)\cr names of each variable/level corresponding to pair of proportions in
192		#' `x` and `y`. Must be of equal length to `x` and `y`.
193		#' @param pct (`flag`)\cr whether output should be returned as percentages. Defaults to `TRUE`.
194		#'
195		#' @return List of proportion differences and CIs corresponding to each pair of number of occurrences in `x` and
196		#' `y`. Each list element consists of 3 statistics: proportion difference, CI lower bound, and CI upper bound.
197		#'
198		#' @seealso Split function [add_riskdiff()] which, when used as `split_fun` within [rtables::split_cols_by()]
199		#' with `riskdiff` argument is set to `TRUE` in subsequent analyze functions, adds a column containing
200		#' proportion (risk) difference to an `rtables` layout.
201		#'
202		#' @examples
203		#' stat_propdiff_ci(
204		#' x = list(0.375), y = list(0.01), N_x = 5, N_y = 5, list_names = "x", conf_level = 0.9
205		#' )
206		#'
207		#' stat_propdiff_ci(
208		#' x = list(0.5, 0.75, 1), y = list(0.25, 0.05, 0.5), N_x = 10, N_y = 20, pct = FALSE
209		#' )
210		#'
211		#' @export
212		stat_propdiff_ci <- function(x,
213		y,
214		N_x, # nolint
215		N_y, # nolint
216		list_names = NULL,
217		conf_level = 0.95,
218		pct = TRUE) {
219	47x	checkmate::assert_list(x, types = "numeric")
220	47x	checkmate::assert_list(y, types = "numeric", len = length(x))
221	47x	checkmate::assert_character(list_names, len = length(x), null.ok = TRUE)
222	47x	rd_list <- lapply(seq_along(x), function(i) {
223	120x	p_x <- x[[i]] / N_x
224	120x	p_y <- y[[i]] / N_y
225	120x	rd_ci <- p_x - p_y + c(-1, 1) * stats::qnorm((1 + conf_level) / 2) *
226	120x	sqrt(p_x * (1 - p_x) / N_x + p_y * (1 - p_y) / N_y)
227	120x	c(p_x - p_y, rd_ci) * ifelse(pct, 100, 1)
228		})
229	47x	names(rd_list) <- list_names
230	47x	rd_list
231		}

1		#' Cumulative counts of numeric variable by thresholds
2		#'
3		#' @description `r lifecycle::badge("stable")`
4		#'
5		#' The analyze function [count_cumulative()] creates a layout element to calculate cumulative counts of values in a
6		#' numeric variable that are less than, less or equal to, greater than, or greater or equal to user-specified
7		#' threshold values.
8		#'
9		#' This function analyzes numeric variable `vars` against the threshold values supplied to the `thresholds`
10		#' argument as a numeric vector. Whether counts should include the threshold values, and whether to count
11		#' values lower or higher than the threshold values can be set via the `include_eq` and `lower_tail`
12		#' parameters, respectively.
13		#'
14		#' @inheritParams h_count_cumulative
15		#' @inheritParams argument_convention
16		#' @param thresholds (`numeric`)\cr vector of cutoff values for the counts.
17		#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("count_cumulative")`
18		#' to see available statistics for this function.
19		#'
20		#' @seealso Relevant helper function [h_count_cumulative()], and descriptive function [d_count_cumulative()].
21		#'
22		#' @name count_cumulative
23		#' @order 1
24		NULL
25
26		#' Helper function for `s_count_cumulative()`
27		#'
28		#' @description `r lifecycle::badge("stable")`
29		#'
30		#' Helper function to calculate count and fraction of `x` values in the lower or upper tail given a threshold.
31		#'
32		#' @inheritParams argument_convention
33		#' @param threshold (`numeric(1)`)\cr a cutoff value as threshold to count values of `x`.
34		#' @param lower_tail (`flag`)\cr whether to count lower tail, default is `TRUE`.
35		#' @param include_eq (`flag`)\cr whether to include value equal to the `threshold` in
36		#' count, default is `TRUE`.
37		#'
38		#' @return A named vector with items:
39		#' * `count`: the count of values less than, less or equal to, greater than, or greater or equal to a threshold
40		#' of user specification.
41		#' * `fraction`: the fraction of the count.
42		#'
43		#' @seealso [count_cumulative]
44		#'
45		#' @examples
46		#' set.seed(1, kind = "Mersenne-Twister")
47		#' x <- c(sample(1:10, 10), NA)
48		#' .N_col <- length(x)
49		#'
50		#' h_count_cumulative(x, 5, .N_col = .N_col)
51		#' h_count_cumulative(x, 5, lower_tail = FALSE, include_eq = FALSE, na.rm = FALSE, .N_col = .N_col)
52		#' h_count_cumulative(x, 0, lower_tail = FALSE, .N_col = .N_col)
53		#' h_count_cumulative(x, 100, lower_tail = FALSE, .N_col = .N_col)
54		#'
55		#' @export
56		h_count_cumulative <- function(x,
57		threshold,
58		lower_tail = TRUE,
59		include_eq = TRUE,
60		na.rm = TRUE, # nolint
61		.N_col) { # nolint
62	20x	checkmate::assert_numeric(x)
63	20x	checkmate::assert_numeric(threshold)
64	20x	checkmate::assert_numeric(.N_col)
65	20x	checkmate::assert_flag(lower_tail)
66	20x	checkmate::assert_flag(include_eq)
67	20x	checkmate::assert_flag(na.rm)
68
69	20x	is_keep <- if (na.rm) !is.na(x) else rep(TRUE, length(x))
70	20x	count <- if (lower_tail && include_eq) {
71	7x	length(x[is_keep & x <= threshold])
72	20x	} else if (lower_tail && !include_eq) {
73	!	length(x[is_keep & x < threshold])
74	20x	} else if (!lower_tail && include_eq) {
75	6x	length(x[is_keep & x >= threshold])
76	20x	} else if (!lower_tail && !include_eq) {
77	7x	length(x[is_keep & x > threshold])
78		}
79
80	20x	result <- c(count = count, fraction = count / .N_col)
81	20x	result
82		}
83
84		#' Description of cumulative count
85		#'
86		#' @description `r lifecycle::badge("stable")`
87		#'
88		#' This is a helper function that describes the analysis in [s_count_cumulative()].
89		#'
90		#' @inheritParams h_count_cumulative
91		#'
92		#' @return Labels for [s_count_cumulative()].
93		#'
94		#' @export
95		d_count_cumulative <- function(threshold, lower_tail = TRUE, include_eq = TRUE) {
96	18x	checkmate::assert_numeric(threshold)
97	18x	lg <- if (lower_tail) "<" else ">"
98	18x	eq <- if (include_eq) "=" else ""
99	18x	paste0(lg, eq, " ", threshold)
100		}
101
102		#' @describeIn count_cumulative Statistics function that produces a named list given a numeric vector of thresholds.
103		#'
104		#' @return
105		#' * `s_count_cumulative()` returns a named list of `count_fraction`s: a list with each `thresholds` value as a
106		#' component, each component containing a vector for the count and fraction.
107		#'
108		#' @keywords internal
109		s_count_cumulative <- function(x,
110		thresholds,
111		lower_tail = TRUE,
112		include_eq = TRUE,
113		.N_col, # nolint
114		...) {
115	5x	checkmate::assert_numeric(thresholds, min.len = 1, any.missing = FALSE)
116
117	5x	count_fraction_list <- Map(function(thres) {
118	10x	result <- h_count_cumulative(x, thres, lower_tail, include_eq, .N_col = .N_col, ...)
119	10x	label <- d_count_cumulative(thres, lower_tail, include_eq)
120	10x	formatters::with_label(result, label)
121	5x	}, thresholds)
122
123	5x	names(count_fraction_list) <- thresholds
124	5x	list(count_fraction = count_fraction_list)
125		}
126
127		#' @describeIn count_cumulative Formatted analysis function which is used as `afun`
128		#' in `count_cumulative()`.
129		#'
130		#' @return
131		#' * `a_count_cumulative()` returns the corresponding list with formatted [rtables::CellValue()].
132		#'
133		#' @keywords internal
134		a_count_cumulative <- make_afun(
135		s_count_cumulative,
136		.formats = c(count_fraction = format_count_fraction)
137		)
138
139		#' @describeIn count_cumulative Layout-creating function which can take statistics function arguments
140		#' and additional format arguments. This function is a wrapper for [rtables::analyze()].
141		#'
142		#' @return
143		#' * `count_cumulative()` returns a layout object suitable for passing to further layouting functions,
144		#' or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
145		#' the statistics from `s_count_cumulative()` to the table layout.
146		#'
147		#' @examples
148		#' basic_table() %>%
149		#' split_cols_by("ARM") %>%
150		#' add_colcounts() %>%
151		#' count_cumulative(
152		#' vars = "AGE",
153		#' thresholds = c(40, 60)
154		#' ) %>%
155		#' build_table(tern_ex_adsl)
156		#'
157		#' @export
158		#' @order 2
159		count_cumulative <- function(lyt,
160		vars,
161		thresholds,
162		lower_tail = TRUE,
163		include_eq = TRUE,
164		var_labels = vars,
165		show_labels = "visible",
166		na_str = default_na_str(),
167		nested = TRUE,
168		...,
169		table_names = vars,
170		.stats = NULL,
171		.formats = NULL,
172		.labels = NULL,
173		.indent_mods = NULL) {
174	2x	extra_args <- list(thresholds = thresholds, lower_tail = lower_tail, include_eq = include_eq, ...)
175
176	2x	afun <- make_afun(
177	2x	a_count_cumulative,
178	2x	.stats = .stats,
179	2x	.formats = .formats,
180	2x	.labels = .labels,
181	2x	.indent_mods = .indent_mods,
182	2x	.ungroup_stats = "count_fraction"
183		)
184	2x	analyze(
185	2x	lyt,
186	2x	vars,
187	2x	afun = afun,
188	2x	na_str = na_str,
189	2x	table_names = table_names,
190	2x	var_labels = var_labels,
191	2x	show_labels = show_labels,
192	2x	nested = nested,
193	2x	extra_args = extra_args
194		)
195		}

1		#' Convert `rtable` objects to `ggplot` objects
2		#'
3		#' @description `r lifecycle::badge("experimental")`
4		#'
5		#' Given a [rtables::rtable()] object, performs basic conversion to a [ggplot2::ggplot()] object built using
6		#' functions from the `ggplot2` package. Any table titles and/or footnotes are ignored.
7		#'
8		#' @param tbl (`VTableTree`)\cr `rtables` table object.
9		#' @param fontsize (`numeric(1)`)\cr font size.
10		#' @param colwidths (`numeric` or `NULL`)\cr a vector of column widths. Each element's position in
11		#' `colwidths` corresponds to the column of `tbl` in the same position. If `NULL`, column widths
12		#' are calculated according to maximum number of characters per column.
13		#' @param lbl_col_padding (`numeric`)\cr additional padding to use when calculating spacing between
14		#' the first (label) column and the second column of `tbl`. If `colwidths` is specified,
15		#' the width of the first column becomes `colwidths[1] + lbl_col_padding`. Defaults to 0.
16		#'
17		#' @return A `ggplot` object.
18		#'
19		#' @examples
20		#' dta <- data.frame(
21		#' ARM = rep(LETTERS[1:3], rep(6, 3)),
22		#' AVISIT = rep(paste0("V", 1:3), 6),
23		#' AVAL = c(9:1, rep(NA, 9))
24		#' )
25		#'
26		#' lyt <- basic_table() %>%
27		#' split_cols_by(var = "ARM") %>%
28		#' split_rows_by(var = "AVISIT") %>%
29		#' analyze_vars(vars = "AVAL")
30		#'
31		#' tbl <- build_table(lyt, df = dta)
32		#'
33		#' rtable2gg(tbl)
34		#'
35		#' rtable2gg(tbl, fontsize = 15, colwidths = c(2, 1, 1, 1))
36		#'
37		#' @export
38		rtable2gg <- function(tbl, fontsize = 12, colwidths = NULL, lbl_col_padding = 0) {
39	6x	mat <- rtables::matrix_form(tbl, indent_rownames = TRUE)
40	6x	mat_strings <- formatters::mf_strings(mat)
41	6x	mat_aligns <- formatters::mf_aligns(mat)
42	6x	mat_indent <- formatters::mf_rinfo(mat)$indent
43	6x	mat_display <- formatters::mf_display(mat)
44	6x	nlines_hdr <- formatters::mf_nlheader(mat)
45	6x	shared_hdr_rows <- which(apply(mat_display, 1, function(x) (any(!x))))
46
47	6x	tbl_df <- data.frame(mat_strings)
48	6x	body_rows <- seq(nlines_hdr + 1, nrow(tbl_df))
49	6x	mat_aligns <- apply(mat_aligns, 1:2, function(x) if (x == "left") 0 else if (x == "right") 1 else 0.5)
50
51		# Apply indentation in first column
52	6x	tbl_df[body_rows, 1] <- sapply(body_rows, function(i) {
53	42x	ind_i <- mat_indent[i - nlines_hdr] * 4
54	18x	if (ind_i > 0) paste0(paste(rep(" ", ind_i), collapse = ""), tbl_df[i, 1]) else tbl_df[i, 1]
55		})
56
57		# Get column widths
58	6x	if (is.null(colwidths)) {
59	6x	colwidths <- apply(tbl_df, 2, function(x) max(nchar(x))) + 1
60		}
61	6x	tot_width <- sum(colwidths) + lbl_col_padding
62
63	6x	if (length(shared_hdr_rows) > 0) {
64	5x	tbl_df <- tbl_df[-shared_hdr_rows, ]
65	5x	mat_aligns <- mat_aligns[-shared_hdr_rows, ]
66		}
67
68	6x	res <- ggplot(data = tbl_df) +
69	6x	theme_void() +
70	6x	scale_x_continuous(limits = c(0, tot_width)) +
71	6x	scale_y_continuous(limits = c(0, nrow(mat_strings))) +
72	6x	annotate(
73	6x	"segment",
74	6x	x = 0, xend = tot_width,
75	6x	y = nrow(mat_strings) - nlines_hdr + 0.5, yend = nrow(mat_strings) - nlines_hdr + 0.5
76		)
77
78		# If header content spans multiple columns, center over these columns
79	6x	if (length(shared_hdr_rows) > 0) {
80	5x	mat_strings[shared_hdr_rows, ] <- trimws(mat_strings[shared_hdr_rows, ])
81	5x	for (hr in shared_hdr_rows) {
82	6x	hdr_lbls <- mat_strings[1:hr, mat_display[hr, -1]]
83	6x	hdr_lbls <- matrix(hdr_lbls[nzchar(hdr_lbls)], nrow = hr)
84	6x	for (idx_hl in seq_len(ncol(hdr_lbls))) {
85	13x	cur_lbl <- tail(hdr_lbls[, idx_hl], 1)
86	13x	which_cols <- if (hr == 1) {
87	9x	which(mat_strings[hr, ] == hdr_lbls[idx_hl])
88	13x	} else { # for >2 col splits, only print labels for each unique combo of nested columns
89	4x	which(
90	4x	apply(mat_strings[1:hr, ], 2, function(x) all(x == hdr_lbls[1:hr, idx_hl]))
91		)
92		}
93	13x	line_pos <- c(
94	13x	sum(colwidths[1:(which_cols[1] - 1)]) + 1 + lbl_col_padding,
95	13x	sum(colwidths[1:max(which_cols)]) - 1 + lbl_col_padding
96		)
97
98	13x	res <- res +
99	13x	annotate(
100	13x	"text",
101	13x	x = mean(line_pos),
102	13x	y = nrow(mat_strings) + 1 - hr,
103	13x	label = cur_lbl,
104	13x	size = fontsize / .pt
105		) +
106	13x	annotate(
107	13x	"segment",
108	13x	x = line_pos[1],
109	13x	xend = line_pos[2],
110	13x	y = nrow(mat_strings) - hr + 0.5,
111	13x	yend = nrow(mat_strings) - hr + 0.5
112		)
113		}
114		}
115		}
116
117		# Add table columns
118	6x	for (i in seq_len(ncol(tbl_df))) {
119	40x	res <- res + annotate(
120	40x	"text",
121	40x	x = if (i == 1) 0 else sum(colwidths[1:i]) - 0.5 * colwidths[i] + lbl_col_padding,
122	40x	y = rev(seq_len(nrow(tbl_df))),
123	40x	label = tbl_df[, i],
124	40x	hjust = mat_aligns[, i],
125	40x	size = fontsize / .pt
126		)
127		}
128
129	6x	res
130		}
131
132		#' Convert `data.frame` object to `ggplot` object
133		#'
134		#' @description `r lifecycle::badge("experimental")`
135		#'
136		#' Given a `data.frame` object, performs basic conversion to a [ggplot2::ggplot()] object built using
137		#' functions from the `ggplot2` package.
138		#'
139		#' @param df (`data.frame`)\cr a data frame.
140		#' @param colwidths (`numeric` or `NULL`)\cr a vector of column widths. Each element's position in
141		#' `colwidths` corresponds to the column of `df` in the same position. If `NULL`, column widths
142		#' are calculated according to maximum number of characters per column.
143		#' @param font_size (`numeric(1)`)\cr font size.
144		#' @param col_labels (`flag`)\cr whether the column names (labels) of `df` should be used as the first row
145		#' of the output table.
146		#' @param col_lab_fontface (`string`)\cr font face to apply to the first row (of column labels
147		#' if `col_labels = TRUE`). Defaults to `"bold"`.
148		#' @param hline (`flag`)\cr whether a horizontal line should be printed below the first row of the table.
149		#' @param bg_fill (`string`)\cr table background fill color.
150		#'
151		#' @return A `ggplot` object.
152		#'
153		#' @examples
154		#' \dontrun{
155		#' df2gg(head(iris, 5))
156		#'
157		#' df2gg(head(iris, 5), font_size = 15, colwidths = c(1, 1, 1, 1, 1))
158		#' }
159		#' @keywords internal
160		df2gg <- function(df,
161		colwidths = NULL,
162		font_size = 10,
163		col_labels = TRUE,
164		col_lab_fontface = "bold",
165		hline = TRUE,
166		bg_fill = NULL) {
167		# convert to text
168	19x	df <- as.data.frame(apply(df, 1:2, function(x) if (is.na(x)) "NA" else as.character(x)))
169
170	19x	if (col_labels) {
171	10x	df <- as.matrix(df)
172	10x	df <- rbind(colnames(df), df)
173		}
174
175		# Get column widths
176	19x	if (is.null(colwidths)) {
177	1x	colwidths <- apply(df, 2, function(x) max(nchar(x), na.rm = TRUE))
178		}
179	19x	tot_width <- sum(colwidths)
180
181	19x	res <- ggplot(data = df) +
182	19x	theme_void() +
183	19x	scale_x_continuous(limits = c(0, tot_width)) +
184	19x	scale_y_continuous(limits = c(1, nrow(df)))
185
186	9x	if (!is.null(bg_fill)) res <- res + theme(plot.background = element_rect(fill = bg_fill))
187
188	19x	if (hline) {
189	10x	res <- res +
190	10x	annotate(
191	10x	"segment",
192	10x	x = 0 + 0.2 * colwidths[2], xend = tot_width - 0.1 * tail(colwidths, 1),
193	10x	y = nrow(df) - 0.5, yend = nrow(df) - 0.5
194		)
195		}
196
197	19x	for (i in seq_len(ncol(df))) {
198	86x	line_pos <- c(
199	86x	if (i == 1) 0 else sum(colwidths[1:(i - 1)]),
200	86x	sum(colwidths[1:i])
201		)
202	86x	res <- res +
203	86x	annotate(
204	86x	"text",
205	86x	x = mean(line_pos),
206	86x	y = rev(seq_len(nrow(df))),
207	86x	label = df[, i],
208	86x	size = font_size / .pt,
209	86x	fontface = if (col_labels) {
210	32x	c(col_lab_fontface, rep("plain", nrow(df) - 1))
211		} else {
212	54x	rep("plain", nrow(df))
213		}
214		)
215		}
216
217	19x	res
218		}

1		#' Occurrence table pruning
2		#'
3		#' @description `r lifecycle::badge("stable")`
4		#'
5		#' Family of constructor and condition functions to flexibly prune occurrence tables.
6		#' The condition functions always return whether the row result is higher than the threshold.
7		#' Since they are of class [CombinationFunction()] they can be logically combined with other condition
8		#' functions.
9		#'
10		#' @note Since most table specifications are worded positively, we name our constructor and condition
11		#' functions positively, too. However, note that the result of [keep_rows()] says what
12		#' should be pruned, to conform with the [rtables::prune_table()] interface.
13		#'
14		#' @examples
15		#' \donttest{
16		#' tab <- basic_table() %>%
17		#' split_cols_by("ARM") %>%
18		#' split_rows_by("RACE") %>%
19		#' split_rows_by("STRATA1") %>%
20		#' summarize_row_groups() %>%
21		#' analyze_vars("COUNTRY", .stats = "count_fraction") %>%
22		#' build_table(DM)
23		#' }
24		#'
25		#' @name prune_occurrences
26		NULL
27
28		#' @describeIn prune_occurrences Constructor for creating pruning functions based on
29		#' a row condition function. This removes all analysis rows (`TableRow`) that should be
30		#' pruned, i.e., don't fulfill the row condition. It removes the sub-tree if there are no
31		#' children left.
32		#'
33		#' @param row_condition (`CombinationFunction`)\cr condition function which works on individual
34		#' analysis rows and flags whether these should be kept in the pruned table.
35		#'
36		#' @return
37		#' * `keep_rows()` returns a pruning function that can be used with [rtables::prune_table()]
38		#' to prune an `rtables` table.
39		#'
40		#' @examples
41		#' \donttest{
42		#' # `keep_rows`
43		#' is_non_empty <- !CombinationFunction(all_zero_or_na)
44		#' prune_table(tab, keep_rows(is_non_empty))
45		#' }
46		#'
47		#' @export
48		keep_rows <- function(row_condition) {
49	6x	checkmate::assert_function(row_condition)
50	6x	function(table_tree) {
51	2256x	if (inherits(table_tree, "TableRow")) {
52	1872x	return(!row_condition(table_tree))
53		}
54	384x	children <- tree_children(table_tree)
55	384x	identical(length(children), 0L)
56		}
57		}
58
59		#' @describeIn prune_occurrences Constructor for creating pruning functions based on
60		#' a condition for the (first) content row in leaf tables. This removes all leaf tables where
61		#' the first content row does not fulfill the condition. It does not check individual rows.
62		#' It then proceeds recursively by removing the sub tree if there are no children left.
63		#'
64		#' @param content_row_condition (`CombinationFunction`)\cr condition function which works on individual
65		#' first content rows of leaf tables and flags whether these leaf tables should be kept in the pruned table.
66		#'
67		#' @return
68		#' * `keep_content_rows()` returns a pruning function that checks the condition on the first content
69		#' row of leaf tables in the table.
70		#'
71		#' @examples
72		#' # `keep_content_rows`
73		#' \donttest{
74		#' more_than_twenty <- has_count_in_cols(atleast = 20L, col_names = names(tab))
75		#' prune_table(tab, keep_content_rows(more_than_twenty))
76		#' }
77		#'
78		#' @export
79		keep_content_rows <- function(content_row_condition) {
80	1x	checkmate::assert_function(content_row_condition)
81	1x	function(table_tree) {
82	166x	if (is_leaf_table(table_tree)) {
83	24x	content_row <- h_content_first_row(table_tree)
84	24x	return(!content_row_condition(content_row))
85		}
86	142x	if (inherits(table_tree, "DataRow")) {
87	120x	return(FALSE)
88		}
89	22x	children <- tree_children(table_tree)
90	22x	identical(length(children), 0L)
91		}
92		}
93
94		#' @describeIn prune_occurrences Constructor for creating condition functions on total counts in the specified columns.
95		#'
96		#' @param atleast (`numeric(1)`)\cr threshold which should be met in order to keep the row.
97		#' @param ... arguments for row or column access, see [`rtables_access`]: either `col_names` (`character`) including
98		#' the names of the columns which should be used, or alternatively `col_indices` (`integer`) giving the indices
99		#' directly instead.
100		#'
101		#' @return
102		#' * `has_count_in_cols()` returns a condition function that sums the counts in the specified column.
103		#'
104		#' @examples
105		#' \donttest{
106		#' more_than_one <- has_count_in_cols(atleast = 1L, col_names = names(tab))
107		#' prune_table(tab, keep_rows(more_than_one))
108		#' }
109		#'
110		#' @export
111		has_count_in_cols <- function(atleast, ...) {
112	6x	checkmate::assert_count(atleast)
113	6x	CombinationFunction(function(table_row) {
114	337x	row_counts <- h_row_counts(table_row, ...)
115	337x	total_count <- sum(row_counts)
116	337x	total_count >= atleast
117		})
118		}
119
120		#' @describeIn prune_occurrences Constructor for creating condition functions on any of the counts in
121		#' the specified columns satisfying a threshold.
122		#'
123		#' @param atleast (`numeric(1)`)\cr threshold which should be met in order to keep the row.
124		#'
125		#' @return
126		#' * `has_count_in_any_col()` returns a condition function that compares the counts in the
127		#' specified columns with the threshold.
128		#'
129		#' @examples
130		#' \donttest{
131		#' # `has_count_in_any_col`
132		#' any_more_than_one <- has_count_in_any_col(atleast = 1L, col_names = names(tab))
133		#' prune_table(tab, keep_rows(any_more_than_one))
134		#' }
135		#'
136		#' @export
137		has_count_in_any_col <- function(atleast, ...) {
138	3x	checkmate::assert_count(atleast)
139	3x	CombinationFunction(function(table_row) {
140	3x	row_counts <- h_row_counts(table_row, ...)
141	3x	any(row_counts >= atleast)
142		})
143		}
144
145		#' @describeIn prune_occurrences Constructor for creating condition functions on total fraction in
146		#' the specified columns.
147		#'
148		#' @return
149		#' * `has_fraction_in_cols()` returns a condition function that sums the counts in the
150		#' specified column, and computes the fraction by dividing by the total column counts.
151		#'
152		#' @examples
153		#' \donttest{
154		#' # `has_fraction_in_cols`
155		#' more_than_five_percent <- has_fraction_in_cols(atleast = 0.05, col_names = names(tab))
156		#' prune_table(tab, keep_rows(more_than_five_percent))
157		#' }
158		#'
159		#' @export
160		has_fraction_in_cols <- function(atleast, ...) {
161	4x	assert_proportion_value(atleast, include_boundaries = TRUE)
162	4x	CombinationFunction(function(table_row) {
163	306x	row_counts <- h_row_counts(table_row, ...)
164	306x	total_count <- sum(row_counts)
165	306x	col_counts <- h_col_counts(table_row, ...)
166	306x	total_n <- sum(col_counts)
167	306x	total_percent <- total_count / total_n
168	306x	total_percent >= atleast
169		})
170		}
171
172		#' @describeIn prune_occurrences Constructor for creating condition functions on any fraction in
173		#' the specified columns.
174		#'
175		#' @return
176		#' * `has_fraction_in_any_col()` returns a condition function that looks at the fractions
177		#' in the specified columns and checks whether any of them fulfill the threshold.
178		#'
179		#' @examples
180		#' \donttest{
181		#' # `has_fraction_in_any_col`
182		#' any_atleast_five_percent <- has_fraction_in_any_col(atleast = 0.05, col_names = names(tab))
183		#' prune_table(tab, keep_rows(any_atleast_five_percent))
184		#' }
185		#'
186		#' @export
187		has_fraction_in_any_col <- function(atleast, ...) {
188	3x	assert_proportion_value(atleast, include_boundaries = TRUE)
189	3x	CombinationFunction(function(table_row) {
190	3x	row_fractions <- h_row_fractions(table_row, ...)
191	3x	any(row_fractions >= atleast)
192		})
193		}
194
195		#' @describeIn prune_occurrences Constructor for creating condition function that checks the difference
196		#' between the fractions reported in each specified column.
197		#'
198		#' @return
199		#' * `has_fractions_difference()` returns a condition function that extracts the fractions of each
200		#' specified column, and computes the difference of the minimum and maximum.
201		#'
202		#' @examples
203		#' \donttest{
204		#' # `has_fractions_difference`
205		#' more_than_five_percent_diff <- has_fractions_difference(atleast = 0.05, col_names = names(tab))
206		#' prune_table(tab, keep_rows(more_than_five_percent_diff))
207		#' }
208		#'
209		#' @export
210		has_fractions_difference <- function(atleast, ...) {
211	4x	assert_proportion_value(atleast, include_boundaries = TRUE)
212	4x	CombinationFunction(function(table_row) {
213	246x	fractions <- h_row_fractions(table_row, ...)
214	246x	difference <- diff(range(fractions))
215	246x	difference >= atleast
216		})
217		}
218
219		#' @describeIn prune_occurrences Constructor for creating condition function that checks the difference
220		#' between the counts reported in each specified column.
221		#'
222		#' @return
223		#' * `has_counts_difference()` returns a condition function that extracts the counts of each
224		#' specified column, and computes the difference of the minimum and maximum.
225		#'
226		#' @examples
227		#' \donttest{
228		#' more_than_one_diff <- has_counts_difference(atleast = 1L, col_names = names(tab))
229		#' prune_table(tab, keep_rows(more_than_one_diff))
230		#' }
231		#'
232		#' @export
233		has_counts_difference <- function(atleast, ...) {
234	4x	checkmate::assert_count(atleast)
235	4x	CombinationFunction(function(table_row) {
236	30x	counts <- h_row_counts(table_row, ...)
237	30x	difference <- diff(range(counts))
238	30x	difference >= atleast
239		})
240		}

1		#' Cox proportional hazards regression
2		#'
3		#' @description `r lifecycle::badge("stable")`
4		#'
5		#' Fits a Cox regression model and estimates hazard ratio to describe the effect size in a survival analysis.
6		#'
7		#' @inheritParams argument_convention
8		#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("summarize_coxreg")`
9		#' to see available statistics for this function.
10		#'
11		#' @details Cox models are the most commonly used methods to estimate the magnitude of
12		#' the effect in survival analysis. It assumes proportional hazards: the ratio
13		#' of the hazards between groups (e.g., two arms) is constant over time.
14		#' This ratio is referred to as the "hazard ratio" (HR) and is one of the
15		#' most commonly reported metrics to describe the effect size in survival
16		#' analysis (NEST Team, 2020).
17		#'
18		#' @seealso [fit_coxreg] for relevant fitting functions, [h_cox_regression] for relevant
19		#' helper functions, and [tidy_coxreg] for custom tidy methods.
20		#'
21		#' @examples
22		#' library(survival)
23		#'
24		#' # Testing dataset [survival::bladder].
25		#' set.seed(1, kind = "Mersenne-Twister")
26		#' dta_bladder <- with(
27		#' data = bladder[bladder$enum < 5, ],
28		#' tibble::tibble(
29		#' TIME = stop,
30		#' STATUS = event,
31		#' ARM = as.factor(rx),
32		#' COVAR1 = as.factor(enum) %>% formatters::with_label("A Covariate Label"),
33		#' COVAR2 = factor(
34		#' sample(as.factor(enum)),
35		#' levels = 1:4, labels = c("F", "F", "M", "M")
36		#' ) %>% formatters::with_label("Sex (F/M)")
37		#' )
38		#' )
39		#' dta_bladder$AGE <- sample(20:60, size = nrow(dta_bladder), replace = TRUE)
40		#' dta_bladder$STUDYID <- factor("X")
41		#'
42		#' u1_variables <- list(
43		#' time = "TIME", event = "STATUS", arm = "ARM", covariates = c("COVAR1", "COVAR2")
44		#' )
45		#'
46		#' u2_variables <- list(time = "TIME", event = "STATUS", covariates = c("COVAR1", "COVAR2"))
47		#'
48		#' m1_variables <- list(
49		#' time = "TIME", event = "STATUS", arm = "ARM", covariates = c("COVAR1", "COVAR2")
50		#' )
51		#'
52		#' m2_variables <- list(time = "TIME", event = "STATUS", covariates = c("COVAR1", "COVAR2"))
53		#'
54		#' @name cox_regression
55		#' @order 1
56		NULL
57
58		#' @describeIn cox_regression Statistics function that transforms results tabulated
59		#' from [fit_coxreg_univar()] or [fit_coxreg_multivar()] into a list.
60		#'
61		#' @param model_df (`data.frame`)\cr contains the resulting model fit from a [fit_coxreg]
62		#' function with tidying applied via [broom::tidy()].
63		#' @param .stats (`character`)\cr the names of statistics to be reported among:
64		#' * `n`: number of observations (univariate only)
65		#' * `hr`: hazard ratio
66		#' * `ci`: confidence interval
67		#' * `pval`: p-value of the treatment effect
68		#' * `pval_inter`: p-value of the interaction effect between the treatment and the covariate (univariate only)
69		#' @param .which_vars (`character`)\cr which rows should statistics be returned for from the given model.
70		#' Defaults to `"all"`. Other options include `"var_main"` for main effects, `"inter"` for interaction effects,
71		#' and `"multi_lvl"` for multivariate model covariate level rows. When `.which_vars` is `"all"`, specific
72		#' variables can be selected by specifying `.var_nms`.
73		#' @param .var_nms (`character`)\cr the `term` value of rows in `df` for which `.stats` should be returned. Typically
74		#' this is the name of a variable. If using variable labels, `var` should be a vector of both the desired
75		#' variable name and the variable label in that order to see all `.stats` related to that variable. When `.which_vars`
76		#' is `"var_main"`, `.var_nms` should be only the variable name.
77		#'
78		#' @return
79		#' * `s_coxreg()` returns the selected statistic for from the Cox regression model for the selected variable(s).
80		#'
81		#' @examples
82		#' # s_coxreg
83		#'
84		#' # Univariate
85		#' univar_model <- fit_coxreg_univar(variables = u1_variables, data = dta_bladder)
86		#' df1 <- broom::tidy(univar_model)
87		#'
88		#' s_coxreg(model_df = df1, .stats = "hr")
89		#'
90		#' # Univariate with interactions
91		#' univar_model_inter <- fit_coxreg_univar(
92		#' variables = u1_variables, control = control_coxreg(interaction = TRUE), data = dta_bladder
93		#' )
94		#' df1_inter <- broom::tidy(univar_model_inter)
95		#'
96		#' s_coxreg(model_df = df1_inter, .stats = "hr", .which_vars = "inter", .var_nms = "COVAR1")
97		#'
98		#' # Univariate without treatment arm - only "COVAR2" covariate effects
99		#' univar_covs_model <- fit_coxreg_univar(variables = u2_variables, data = dta_bladder)
100		#' df1_covs <- broom::tidy(univar_covs_model)
101		#'
102		#' s_coxreg(model_df = df1_covs, .stats = "hr", .var_nms = c("COVAR2", "Sex (F/M)"))
103		#'
104		#' # Multivariate.
105		#' multivar_model <- fit_coxreg_multivar(variables = m1_variables, data = dta_bladder)
106		#' df2 <- broom::tidy(multivar_model)
107		#'
108		#' s_coxreg(model_df = df2, .stats = "pval", .which_vars = "var_main", .var_nms = "COVAR1")
109		#' s_coxreg(
110		#' model_df = df2, .stats = "pval", .which_vars = "multi_lvl",
111		#' .var_nms = c("COVAR1", "A Covariate Label")
112		#' )
113		#'
114		#' # Multivariate without treatment arm - only "COVAR1" main effect
115		#' multivar_covs_model <- fit_coxreg_multivar(variables = m2_variables, data = dta_bladder)
116		#' df2_covs <- broom::tidy(multivar_covs_model)
117		#'
118		#' s_coxreg(model_df = df2_covs, .stats = "hr")
119		#'
120		#' @export
121		s_coxreg <- function(model_df, .stats, .which_vars = "all", .var_nms = NULL) {
122	291x	assert_df_with_variables(model_df, list(term = "term", stat = .stats))
123	291x	checkmate::assert_multi_class(model_df$term, classes = c("factor", "character"))
124	291x	model_df$term <- as.character(model_df$term)
125	291x	.var_nms <- .var_nms[!is.na(.var_nms)]
126
127	289x	if (length(.var_nms) > 0) model_df <- model_df[model_df$term %in% .var_nms, ]
128	69x	if (.which_vars == "multi_lvl") model_df$term <- tail(.var_nms, 1)
129
130		# We need a list with names corresponding to the stats to display of equal length to the list of stats.
131	291x	y <- split(model_df, f = model_df$term, drop = FALSE)
132	291x	y <- stats::setNames(y, nm = rep(.stats, length(y)))
133
134	291x	if (.which_vars == "var_main") {
135	128x	y <- lapply(y, function(x) x[1, ]) # only main effect
136	163x	} else if (.which_vars %in% c("inter", "multi_lvl")) {
137	120x	y <- lapply(y, function(x) if (nrow(y[[1]]) > 1) x[-1, ] else x) # exclude main effect
138		}
139
140	291x	lapply(
141	291x	X = y,
142	291x	FUN = function(x) {
143	295x	z <- as.list(x[[.stats]])
144	295x	stats::setNames(z, nm = x$term_label)
145		}
146		)
147		}
148
149		#' @describeIn cox_regression Analysis function which is used as `afun` in [rtables::analyze()]
150		#' and `cfun` in [rtables::summarize_row_groups()] within `summarize_coxreg()`.
151		#'
152		#' @param eff (`flag`)\cr whether treatment effect should be calculated. Defaults to `FALSE`.
153		#' @param var_main (`flag`)\cr whether main effects should be calculated. Defaults to `FALSE`.
154		#' @param na_str (`string`)\cr custom string to replace all `NA` values with. Defaults to `""`.
155		#' @param cache_env (`environment`)\cr an environment object used to cache the regression model in order to
156		#' avoid repeatedly fitting the same model for every row in the table. Defaults to `NULL` (no caching).
157		#' @param varlabels (`list`)\cr a named list corresponds to the names of variables found in data, passed
158		#' as a named list and corresponding to time, event, arm, strata, and covariates terms. If arm is missing
159		#' from variables, then only Cox model(s) including the covariates will be fitted and the corresponding
160		#' effect estimates will be tabulated later.
161		#'
162		#' @return
163		#' * `a_coxreg()` returns formatted [rtables::CellValue()].
164		#'
165		#' @examples
166		#' a_coxreg(
167		#' df = dta_bladder,
168		#' labelstr = "Label 1",
169		#' variables = u1_variables,
170		#' .spl_context = list(value = "COVAR1"),
171		#' .stats = "n",
172		#' .formats = "xx"
173		#' )
174		#'
175		#' a_coxreg(
176		#' df = dta_bladder,
177		#' labelstr = "",
178		#' variables = u1_variables,
179		#' .spl_context = list(value = "COVAR2"),
180		#' .stats = "pval",
181		#' .formats = "xx.xxxx"
182		#' )
183		#'
184		#' @export
185		a_coxreg <- function(df,
186		labelstr,
187		eff = FALSE,
188		var_main = FALSE,
189		multivar = FALSE,
190		variables,
191		at = list(),
192		control = control_coxreg(),
193		.spl_context,
194		.stats,
195		.formats,
196		.indent_mods = NULL,
197		na_str = "",
198		cache_env = NULL) {
199	288x	cov_no_arm <- !multivar && !"arm" %in% names(variables) && control$interaction # special case: univar no arm
200	288x	cov <- tail(.spl_context$value, 1) # current variable/covariate
201	288x	var_lbl <- formatters::var_labels(df)[cov] # check for df labels
202	288x	if (length(labelstr) > 1) {
203	8x	labelstr <- if (cov %in% names(labelstr)) labelstr[[cov]] else var_lbl # use df labels if none
204	280x	} else if (!is.na(var_lbl) && labelstr == cov && cov %in% variables$covariates) {
205	67x	labelstr <- var_lbl
206		}
207	288x	if (eff \|\| multivar \|\| cov_no_arm) {
208	143x	control$interaction <- FALSE
209		} else {
210	145x	variables$covariates <- cov
211	50x	if (var_main) control$interaction <- TRUE
212		}
213
214	288x	if (is.null(cache_env[[cov]])) {
215	47x	if (!multivar) {
216	32x	model <- fit_coxreg_univar(variables = variables, data = df, at = at, control = control) %>% broom::tidy()
217		} else {
218	15x	model <- fit_coxreg_multivar(variables = variables, data = df, control = control) %>% broom::tidy()
219		}
220	47x	cache_env[[cov]] <- model
221		} else {
222	241x	model <- cache_env[[cov]]
223		}
224	148x	if (!multivar && !var_main) model[, "pval_inter"] <- NA_real_
225
226	288x	if (cov_no_arm \|\| (!cov_no_arm && !"arm" %in% names(variables) && is.numeric(df[[cov]]))) {
227	15x	multivar <- TRUE
228	3x	if (!cov_no_arm) var_main <- TRUE
229		}
230
231	288x	vars_coxreg <- list(which_vars = "all", var_nms = NULL)
232	288x	if (eff) {
233	65x	if (multivar && !var_main) { # multivar treatment level
234	12x	var_lbl_arm <- formatters::var_labels(df)[[variables$arm]]
235	12x	vars_coxreg[c("var_nms", "which_vars")] <- list(c(variables$arm, var_lbl_arm), "multi_lvl")
236		} else { # treatment effect
237	53x	vars_coxreg["var_nms"] <- variables$arm
238	12x	if (var_main) vars_coxreg["which_vars"] <- "var_main"
239		}
240		} else {
241	223x	if (!multivar \|\| (multivar && var_main && !is.numeric(df[[cov]]))) { # covariate effect/level
242	166x	vars_coxreg[c("var_nms", "which_vars")] <- list(cov, "var_main")
243	57x	} else if (multivar) { # multivar covariate level
244	57x	vars_coxreg[c("var_nms", "which_vars")] <- list(c(cov, var_lbl), "multi_lvl")
245	12x	if (var_main) model[cov, .stats] <- NA_real_
246		}
247	50x	if (!multivar && !var_main && control$interaction) vars_coxreg["which_vars"] <- "inter" # interaction effect
248		}
249	288x	var_vals <- s_coxreg(model, .stats, .which_vars = vars_coxreg$which_vars, .var_nms = vars_coxreg$var_nms)[[1]]
250	288x	var_names <- if (all(grepl("\\(reference = ", names(var_vals))) && labelstr != tail(.spl_context$value, 1)) {
251	27x	paste(c(labelstr, tail(strsplit(names(var_vals), " ")[[1]], 3)), collapse = " ") # "reference" main effect labels
252	288x	} else if ((!multivar && !eff && !(!var_main && control$interaction) && nchar(labelstr) > 0) \|\|
253	288x	(multivar && var_main && is.numeric(df[[cov]]))) { # nolint
254	71x	labelstr # other main effect labels
255	288x	} else if (multivar && !eff && !var_main && is.numeric(df[[cov]])) {
256	12x	"All" # multivar numeric covariate
257		} else {
258	178x	names(var_vals)
259		}
260	288x	in_rows(
261	288x	.list = var_vals, .names = var_names, .labels = var_names, .indent_mods = .indent_mods,
262	288x	.formats = stats::setNames(rep(.formats, length(var_names)), var_names),
263	288x	.format_na_strs = stats::setNames(rep(na_str, length(var_names)), var_names)
264		)
265		}
266
267		#' @describeIn cox_regression Layout-creating function which creates a Cox regression summary table
268		#' layout. This function is a wrapper for several `rtables` layouting functions. This function
269		#' is a wrapper for [rtables::analyze_colvars()] and [rtables::summarize_row_groups()].
270		#'
271		#' @inheritParams fit_coxreg_univar
272		#' @param multivar (`flag`)\cr whether multivariate Cox regression should run (defaults to `FALSE`), otherwise
273		#' univariate Cox regression will run.
274		#' @param common_var (`string`)\cr the name of a factor variable in the dataset which takes the same value
275		#' for all rows. This should be created during pre-processing if no such variable currently exists.
276		#' @param .section_div (`string` or `NA`)\cr string which should be repeated as a section divider between sections.
277		#' Defaults to `NA` for no section divider. If a vector of two strings are given, the first will be used between
278		#' treatment and covariate sections and the second between different covariates.
279		#'
280		#' @return
281		#' * `summarize_coxreg()` returns a layout object suitable for passing to further layouting functions,
282		#' or to [rtables::build_table()]. Adding this function to an `rtable` layout will add a Cox regression table
283		#' containing the chosen statistics to the table layout.
284		#'
285		#' @seealso [fit_coxreg_univar()] and [fit_coxreg_multivar()] which also take the `variables`, `data`,
286		#' `at` (univariate only), and `control` arguments but return unformatted univariate and multivariate
287		#' Cox regression models, respectively.
288		#'
289		#' @examples
290		#' # summarize_coxreg
291		#'
292		#' result_univar <- basic_table() %>%
293		#' summarize_coxreg(variables = u1_variables) %>%
294		#' build_table(dta_bladder)
295		#' result_univar
296		#'
297		#' result_univar_covs <- basic_table() %>%
298		#' summarize_coxreg(
299		#' variables = u2_variables,
300		#' ) %>%
301		#' build_table(dta_bladder)
302		#' result_univar_covs
303		#'
304		#' result_multivar <- basic_table() %>%
305		#' summarize_coxreg(
306		#' variables = m1_variables,
307		#' multivar = TRUE,
308		#' ) %>%
309		#' build_table(dta_bladder)
310		#' result_multivar
311		#'
312		#' result_multivar_covs <- basic_table() %>%
313		#' summarize_coxreg(
314		#' variables = m2_variables,
315		#' multivar = TRUE,
316		#' varlabels = c("Covariate 1", "Covariate 2") # custom labels
317		#' ) %>%
318		#' build_table(dta_bladder)
319		#' result_multivar_covs
320		#'
321		#' @export
322		#' @order 2
323		summarize_coxreg <- function(lyt,
324		variables,
325		control = control_coxreg(),
326		at = list(),
327		multivar = FALSE,
328		common_var = "STUDYID",
329		.stats = c("n", "hr", "ci", "pval", "pval_inter"),
330		.formats = c(
331		n = "xx", hr = "xx.xx", ci = "(xx.xx, xx.xx)",
332		pval = "x.xxxx \| (<0.0001)", pval_inter = "x.xxxx \| (<0.0001)"
333		),
334		varlabels = NULL,
335		.indent_mods = NULL,
336		na_str = "",
337		.section_div = NA_character_) {
338	16x	if (multivar && control$interaction) {
339	1x	warning(paste(
340	1x	"Interactions are not available for multivariate cox regression using summarize_coxreg.",
341	1x	"The model will be calculated without interaction effects."
342		))
343		}
344	16x	if (control$interaction && !"arm" %in% names(variables)) {
345	1x	stop("To include interactions please specify 'arm' in variables.")
346		}
347
348	15x	.stats <- if (!"arm" %in% names(variables) \|\| multivar) { # only valid statistics
349	6x	intersect(c("hr", "ci", "pval"), .stats)
350	15x	} else if (control$interaction) {
351	5x	intersect(c("n", "hr", "ci", "pval", "pval_inter"), .stats)
352		} else {
353	4x	intersect(c("n", "hr", "ci", "pval"), .stats)
354		}
355	15x	stat_labels <- c(
356	15x	n = "n", hr = "Hazard Ratio", ci = paste0(control$conf_level * 100, "% CI"),
357	15x	pval = "p-value", pval_inter = "Interaction p-value"
358		)
359	15x	stat_labels <- stat_labels[names(stat_labels) %in% .stats]
360	15x	.formats <- .formats[names(.formats) %in% .stats]
361	15x	env <- new.env() # create caching environment
362
363	15x	lyt <- lyt %>%
364	15x	split_cols_by_multivar(
365	15x	vars = rep(common_var, length(.stats)),
366	15x	varlabels = stat_labels,
367	15x	extra_args = list(
368	15x	.stats = .stats, .formats = .formats, .indent_mods = .indent_mods, na_str = rep(na_str, length(.stats)),
369	15x	cache_env = replicate(length(.stats), list(env))
370		)
371		)
372
373	15x	if ("arm" %in% names(variables)) { # treatment effect
374	13x	lyt <- lyt %>%
375	13x	split_rows_by(
376	13x	common_var,
377	13x	split_label = "Treatment:",
378	13x	label_pos = "visible",
379	13x	child_labels = "hidden",
380	13x	section_div = head(.section_div, 1)
381		)
382	13x	if (!multivar) {
383	9x	lyt <- lyt %>%
384	9x	analyze_colvars(
385	9x	afun = a_coxreg,
386	9x	na_str = na_str,
387	9x	extra_args = list(
388	9x	variables = variables, control = control, multivar = multivar, eff = TRUE, var_main = multivar,
389	9x	labelstr = ""
390		)
391		)
392		} else { # treatment level effects
393	4x	lyt <- lyt %>%
394	4x	summarize_row_groups(
395	4x	cfun = a_coxreg,
396	4x	na_str = na_str,
397	4x	extra_args = list(
398	4x	variables = variables, control = control, multivar = multivar, eff = TRUE, var_main = multivar
399		)
400		) %>%
401	4x	analyze_colvars(
402	4x	afun = a_coxreg,
403	4x	na_str = na_str,
404	4x	extra_args = list(eff = TRUE, control = control, variables = variables, multivar = multivar, labelstr = "")
405		)
406		}
407		}
408
409	15x	if ("covariates" %in% names(variables)) { # covariate main effects
410	15x	lyt <- lyt %>%
411	15x	split_rows_by_multivar(
412	15x	vars = variables$covariates,
413	15x	varlabels = varlabels,
414	15x	split_label = "Covariate:",
415	15x	nested = FALSE,
416	15x	child_labels = if (multivar \|\| control$interaction \|\| !"arm" %in% names(variables)) "default" else "hidden",
417	15x	section_div = tail(.section_div, 1)
418		)
419	15x	if (multivar \|\| control$interaction \|\| !"arm" %in% names(variables)) {
420	11x	lyt <- lyt %>%
421	11x	summarize_row_groups(
422	11x	cfun = a_coxreg,
423	11x	na_str = na_str,
424	11x	extra_args = list(
425	11x	variables = variables, at = at, control = control, multivar = multivar,
426	11x	var_main = if (multivar) multivar else control$interaction
427		)
428		)
429		} else {
430	1x	if (!is.null(varlabels)) names(varlabels) <- variables$covariates
431	4x	lyt <- lyt %>%
432	4x	analyze_colvars(
433	4x	afun = a_coxreg,
434	4x	na_str = na_str,
435	4x	extra_args = list(
436	4x	variables = variables, at = at, control = control, multivar = multivar,
437	4x	var_main = if (multivar) multivar else control$interaction,
438	4x	labelstr = if (is.null(varlabels)) "" else varlabels
439		)
440		)
441		}
442
443	2x	if (!"arm" %in% names(variables)) control$interaction <- TRUE # special case: univar no arm
444	15x	if (multivar \|\| control$interaction) { # covariate level effects
445	11x	lyt <- lyt %>%
446	11x	analyze_colvars(
447	11x	afun = a_coxreg,
448	11x	na_str = na_str,
449	11x	extra_args = list(variables = variables, at = at, control = control, multivar = multivar, labelstr = ""),
450	11x	indent_mod = if (!"arm" %in% names(variables) \|\| multivar) 0L else -1L
451		)
452		}
453		}
454
455	15x	lyt
456		}

1		#' Count patients with abnormal analysis range values by baseline status
2		#'
3		#' @description `r lifecycle::badge("stable")`
4		#'
5		#' The analyze function [count_abnormal_by_baseline()] creates a layout element to count patients with abnormal
6		#' analysis range values, categorized by baseline status.
7		#'
8		#' This function analyzes primary analysis variable `var` which indicates abnormal range results. Additional
9		#' analysis variables that can be supplied as a list via the `variables` parameter are `id` (defaults to
10		#' `USUBJID`), a variable to indicate unique subject identifiers, and `baseline` (defaults to `BNRIND`), a
11		#' variable to indicate baseline reference ranges.
12		#'
13		#' For each direction specified via the `abnormal` parameter (e.g. High or Low), we condition on baseline
14		#' range result and count patients in the numerator and denominator as follows for each of the following
15		#' categories:
16		#' * `Not <abnormality>`
17		#' * `num`: The number of patients without abnormality at baseline (excluding those with missing baseline)
18		#' and with at least one abnormality post-baseline.
19		#' * `denom`: The number of patients without abnormality at baseline (excluding those with missing baseline).
20		#' * `<Abnormality>`
21		#' * `num`: The number of patients with abnormality as baseline and at least one abnormality post-baseline.
22		#' * `denom`: The number of patients with abnormality at baseline.
23		#' * `Total`
24		#' * `num`: The number of patients with at least one post-baseline record and at least one abnormality
25		#' post-baseline.
26		#' * `denom`: The number of patients with at least one post-baseline record.
27		#'
28		#' This function assumes that `df` has been filtered to only include post-baseline records.
29		#'
30		#' @inheritParams argument_convention
31		#' @param abnormal (`character`)\cr values identifying the abnormal range level(s) in `.var`.
32		#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("abnormal_by_baseline")`
33		#' to see available statistics for this function.
34		#'
35		#' @note
36		#' * `df` should be filtered to include only post-baseline records.
37		#' * If the baseline variable or analysis variable contains `NA` records, it is expected that `df` has been
38		#' pre-processed using [df_explicit_na()] or [explicit_na()].
39		#'
40		#' @seealso Relevant description function [d_count_abnormal_by_baseline()].
41		#'
42		#' @name abnormal_by_baseline
43		#' @order 1
44		NULL
45
46		#' Description function for `s_count_abnormal_by_baseline()`
47		#'
48		#' @description `r lifecycle::badge("stable")`
49		#'
50		#' Description function that produces the labels for [s_count_abnormal_by_baseline()].
51		#'
52		#' @inheritParams abnormal_by_baseline
53		#'
54		#' @return Abnormal category labels for [s_count_abnormal_by_baseline()].
55		#'
56		#' @examples
57		#' d_count_abnormal_by_baseline("LOW")
58		#'
59		#' @export
60		d_count_abnormal_by_baseline <- function(abnormal) {
61	9x	not_abn_name <- paste("Not", tolower(abnormal))
62	9x	abn_name <- paste0(toupper(substr(abnormal, 1, 1)), tolower(substring(abnormal, 2)))
63	9x	total_name <- "Total"
64
65	9x	list(
66	9x	not_abnormal = not_abn_name,
67	9x	abnormal = abn_name,
68	9x	total = total_name
69		)
70		}
71
72		#' @describeIn abnormal_by_baseline Statistics function for a single `abnormal` level.
73		#'
74		#' @param na_str (`string`)\cr the explicit `na_level` argument you used in the pre-processing steps (maybe with
75		#' [df_explicit_na()]). The default is `"<Missing>"`.
76		#'
77		#' @return
78		#' * `s_count_abnormal_by_baseline()` returns statistic `fraction` which is a named list with 3 labeled elements:
79		#' `not_abnormal`, `abnormal`, and `total`. Each element contains a vector with `num` and `denom` patient counts.
80		#'
81		#' @keywords internal
82		s_count_abnormal_by_baseline <- function(df,
83		.var,
84		abnormal,
85		na_str = "<Missing>",
86		variables = list(id = "USUBJID", baseline = "BNRIND")) {
87	7x	checkmate::assert_string(.var)
88	7x	checkmate::assert_string(abnormal)
89	7x	checkmate::assert_string(na_str)
90	7x	assert_df_with_variables(df, c(range = .var, variables))
91	7x	checkmate::assert_subset(names(variables), c("id", "baseline"))
92	7x	checkmate::assert_multi_class(df[[variables$id]], classes = c("factor", "character"))
93	7x	checkmate::assert_multi_class(df[[variables$baseline]], classes = c("factor", "character"))
94	7x	checkmate::assert_multi_class(df[[.var]], classes = c("factor", "character"))
95
96		# If input is passed as character, changed to factor
97	7x	df[[.var]] <- as_factor_keep_attributes(df[[.var]], na_level = na_str)
98	7x	df[[variables$baseline]] <- as_factor_keep_attributes(df[[variables$baseline]], na_level = na_str)
99
100	7x	assert_valid_factor(df[[.var]], any.missing = FALSE)
101	6x	assert_valid_factor(df[[variables$baseline]], any.missing = FALSE)
102
103		# Keep only records with valid analysis value.
104	5x	df <- df[df[[.var]] != na_str, ]
105
106	5x	anl <- data.frame(
107	5x	id = df[[variables$id]],
108	5x	var = df[[.var]],
109	5x	baseline = df[[variables$baseline]],
110	5x	stringsAsFactors = FALSE
111		)
112
113		# Total:
114		# - Patients in denominator: have at least one valid measurement post-baseline.
115		# - Patients in numerator: have at least one abnormality.
116	5x	total_denom <- length(unique(anl$id))
117	5x	total_num <- length(unique(anl$id[anl$var == abnormal]))
118
119		# Baseline NA records are counted only in total rows.
120	5x	anl <- anl[anl$baseline != na_str, ]
121
122		# Abnormal:
123		# - Patients in denominator: have abnormality at baseline.
124		# - Patients in numerator: have abnormality at baseline AND
125		# have at least one abnormality post-baseline.
126	5x	abn_denom <- length(unique(anl$id[anl$baseline == abnormal]))
127	5x	abn_num <- length(unique(anl$id[anl$baseline == abnormal & anl$var == abnormal]))
128
129		# Not abnormal:
130		# - Patients in denominator: do not have abnormality at baseline.
131		# - Patients in numerator: do not have abnormality at baseline AND
132		# have at least one abnormality post-baseline.
133	5x	not_abn_denom <- length(unique(anl$id[anl$baseline != abnormal]))
134	5x	not_abn_num <- length(unique(anl$id[anl$baseline != abnormal & anl$var == abnormal]))
135
136	5x	labels <- d_count_abnormal_by_baseline(abnormal)
137	5x	list(fraction = list(
138	5x	not_abnormal = formatters::with_label(c(num = not_abn_num, denom = not_abn_denom), labels$not_abnormal),
139	5x	abnormal = formatters::with_label(c(num = abn_num, denom = abn_denom), labels$abnormal),
140	5x	total = formatters::with_label(c(num = total_num, denom = total_denom), labels$total)
141		))
142		}
143
144		#' @describeIn abnormal_by_baseline Formatted analysis function which is used as `afun`
145		#' in `count_abnormal_by_baseline()`.
146		#'
147		#' @return
148		#' * `a_count_abnormal_by_baseline()` returns the corresponding list with formatted [rtables::CellValue()].
149		#'
150		#' @keywords internal
151		a_count_abnormal_by_baseline <- make_afun(
152		s_count_abnormal_by_baseline,
153		.formats = c(fraction = format_fraction)
154		)
155
156		#' @describeIn abnormal_by_baseline Layout-creating function which can take statistics function arguments
157		#' and additional format arguments. This function is a wrapper for [rtables::analyze()].
158		#'
159		#' @return
160		#' * `count_abnormal_by_baseline()` returns a layout object suitable for passing to further layouting functions,
161		#' or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
162		#' the statistics from `s_count_abnormal_by_baseline()` to the table layout.
163		#'
164		#' @examples
165		#' df <- data.frame(
166		#' USUBJID = as.character(c(1:6)),
167		#' ANRIND = factor(c(rep("LOW", 4), "NORMAL", "HIGH")),
168		#' BNRIND = factor(c("LOW", "NORMAL", "HIGH", NA, "LOW", "NORMAL"))
169		#' )
170		#' df <- df_explicit_na(df)
171		#'
172		#' # Layout creating function.
173		#' basic_table() %>%
174		#' count_abnormal_by_baseline(var = "ANRIND", abnormal = c(High = "HIGH")) %>%
175		#' build_table(df)
176		#'
177		#' # Passing of statistics function and formatting arguments.
178		#' df2 <- data.frame(
179		#' ID = as.character(c(1, 2, 3, 4)),
180		#' RANGE = factor(c("NORMAL", "LOW", "HIGH", "HIGH")),
181		#' BLRANGE = factor(c("LOW", "HIGH", "HIGH", "NORMAL"))
182		#' )
183		#'
184		#' basic_table() %>%
185		#' count_abnormal_by_baseline(
186		#' var = "RANGE",
187		#' abnormal = c(Low = "LOW"),
188		#' variables = list(id = "ID", baseline = "BLRANGE"),
189		#' .formats = c(fraction = "xx / xx"),
190		#' .indent_mods = c(fraction = 2L)
191		#' ) %>%
192		#' build_table(df2)
193		#'
194		#' @export
195		#' @order 2
196		count_abnormal_by_baseline <- function(lyt,
197		var,
198		abnormal,
199		variables = list(id = "USUBJID", baseline = "BNRIND"),
200		na_str = "<Missing>",
201		nested = TRUE,
202		...,
203		table_names = abnormal,
204		.stats = NULL,
205		.formats = NULL,
206		.labels = NULL,
207		.indent_mods = NULL) {
208	2x	checkmate::assert_character(abnormal, len = length(table_names), names = "named")
209	2x	checkmate::assert_string(var)
210
211	2x	extra_args <- list(abnormal = abnormal, variables = variables, na_str = na_str, ...)
212
213	2x	afun <- make_afun(
214	2x	a_count_abnormal_by_baseline,
215	2x	.stats = .stats,
216	2x	.formats = .formats,
217	2x	.labels = .labels,
218	2x	.indent_mods = .indent_mods,
219	2x	.ungroup_stats = "fraction"
220		)
221	2x	for (i in seq_along(abnormal)) {
222	4x	extra_args[["abnormal"]] <- abnormal[i]
223
224	4x	lyt <- analyze(
225	4x	lyt = lyt,
226	4x	vars = var,
227	4x	var_labels = names(abnormal[i]),
228	4x	afun = afun,
229	4x	na_str = na_str,
230	4x	nested = nested,
231	4x	table_names = table_names[i],
232	4x	extra_args = extra_args,
233	4x	show_labels = "visible"
234		)
235		}
236	2x	lyt
237		}

1		#' Add titles, footnotes, page Number, and a bounding box to a grid grob
2		#'
3		#' @description `r lifecycle::badge("stable")`
4		#'
5		#' This function is useful to label grid grobs (also `ggplot2`, and `lattice` plots)
6		#' with title, footnote, and page numbers.
7		#'
8		#' @inheritParams grid::grob
9		#' @param grob (`grob`)\cr a grid grob object, optionally `NULL` if only a `grob` with the decoration should be shown.
10		#' @param titles (`character`)\cr titles given as a vector of strings that are each separated by a newline and wrapped
11		#' according to the page width.
12		#' @param footnotes (`character`)\cr footnotes. Uses the same formatting rules as `titles`.
13		#' @param page (`string` or `NULL`)\cr page numeration. If `NULL` then no page number is displayed.
14		#' @param width_titles (`grid::unit`)\cr width of titles. Usually defined as all the available space
15		#' `grid::unit(1, "npc")`, it is affected by the parameter `outer_margins`. Right margins (`outer_margins[4]`)
16		#' need to be subtracted to the allowed width.
17		#' @param width_footnotes (`grid::unit`)\cr width of footnotes. Same default and margin correction as `width_titles`.
18		#' @param border (`flag`)\cr whether a border should be drawn around the plot or not.
19		#' @param padding (`grid::unit`)\cr padding. A unit object of length 4. Innermost margin between the plot (`grob`)
20		#' and, possibly, the border of the plot. Usually expressed in 4 identical values (usually `"lines"`). It defaults
21		#' to `grid::unit(rep(1, 4), "lines")`.
22		#' @param margins (`grid::unit`)\cr margins. A unit object of length 4. Margins between the plot and the other
23		#' elements in the list (e.g. titles, plot, and footers). This is usually expressed in 4 `"lines"`, where the
24		#' lateral ones are 0s, while top and bottom are 1s. It defaults to `grid::unit(c(1, 0, 1, 0), "lines")`.
25		#' @param outer_margins (`grid::unit`)\cr outer margins. A unit object of length 4. It defines the general margin of
26		#' the plot, considering also decorations like titles, footnotes, and page numbers. It defaults to
27		#' `grid::unit(c(2, 1.5, 3, 1.5), "cm")`.
28		#' @param gp_titles (`gpar`)\cr a `gpar` object. Mainly used to set different `"fontsize"`.
29		#' @param gp_footnotes (`gpar`)\cr a `gpar` object. Mainly used to set different `"fontsize"`.
30		#'
31		#' @return A grid grob (`gTree`).
32		#'
33		#' @details The titles and footnotes will be ragged, i.e. each title will be wrapped individually.
34		#'
35		#' @examples
36		#' library(grid)
37		#'
38		#' titles <- c(
39		#' "Edgar Anderson's Iris Data",
40		#' paste(
41		#' "This famous (Fisher's or Anderson's) iris data set gives the measurements",
42		#' "in centimeters of the variables sepal length and width and petal length",
43		#' "and width, respectively, for 50 flowers from each of 3 species of iris."
44		#' )
45		#' )
46		#'
47		#' footnotes <- c(
48		#' "The species are Iris setosa, versicolor, and virginica.",
49		#' paste(
50		#' "iris is a data frame with 150 cases (rows) and 5 variables (columns) named",
51		#' "Sepal.Length, Sepal.Width, Petal.Length, Petal.Width, and Species."
52		#' )
53		#' )
54		#'
55		#' ## empty plot
56		#' grid.newpage()
57		#'
58		#' grid.draw(
59		#' decorate_grob(
60		#' NULL,
61		#' titles = titles,
62		#' footnotes = footnotes,
63		#' page = "Page 4 of 10"
64		#' )
65		#' )
66		#'
67		#' # grid
68		#' p <- gTree(
69		#' children = gList(
70		#' rectGrob(),
71		#' xaxisGrob(),
72		#' yaxisGrob(),
73		#' textGrob("Sepal.Length", y = unit(-4, "lines")),
74		#' textGrob("Petal.Length", x = unit(-3.5, "lines"), rot = 90),
75		#' pointsGrob(iris$Sepal.Length, iris$Petal.Length, gp = gpar(col = iris$Species), pch = 16)
76		#' ),
77		#' vp = vpStack(plotViewport(), dataViewport(xData = iris$Sepal.Length, yData = iris$Petal.Length))
78		#' )
79		#' grid.newpage()
80		#' grid.draw(p)
81		#'
82		#' grid.newpage()
83		#' grid.draw(
84		#' decorate_grob(
85		#' grob = p,
86		#' titles = titles,
87		#' footnotes = footnotes,
88		#' page = "Page 6 of 129"
89		#' )
90		#' )
91		#'
92		#' ## with ggplot2
93		#' library(ggplot2)
94		#'
95		#' p_gg <- ggplot2::ggplot(iris, aes(Sepal.Length, Sepal.Width, col = Species)) +
96		#' ggplot2::geom_point()
97		#' p_gg
98		#' p <- ggplotGrob(p_gg)
99		#' grid.newpage()
100		#' grid.draw(
101		#' decorate_grob(
102		#' grob = p,
103		#' titles = titles,
104		#' footnotes = footnotes,
105		#' page = "Page 6 of 129"
106		#' )
107		#' )
108		#'
109		#' ## with lattice
110		#' library(lattice)
111		#'
112		#' xyplot(Sepal.Length ~ Petal.Length, data = iris, col = iris$Species)
113		#' p <- grid.grab()
114		#' grid.newpage()
115		#' grid.draw(
116		#' decorate_grob(
117		#' grob = p,
118		#' titles = titles,
119		#' footnotes = footnotes,
120		#' page = "Page 6 of 129"
121		#' )
122		#' )
123		#'
124		#' # with gridExtra - no borders
125		#' library(gridExtra)
126		#' grid.newpage()
127		#' grid.draw(
128		#' decorate_grob(
129		#' tableGrob(
130		#' head(mtcars)
131		#' ),
132		#' titles = "title",
133		#' footnotes = "footnote",
134		#' border = FALSE
135		#' )
136		#' )
137		#'
138		#' @export
139		decorate_grob <- function(grob,
140		titles,
141		footnotes,
142		page = "",
143		width_titles = grid::unit(1, "npc"),
144		width_footnotes = grid::unit(1, "npc"),
145		border = TRUE,
146		padding = grid::unit(rep(1, 4), "lines"),
147		margins = grid::unit(c(1, 0, 1, 0), "lines"),
148		outer_margins = grid::unit(c(2, 1.5, 3, 1.5), "cm"),
149		gp_titles = grid::gpar(),
150		gp_footnotes = grid::gpar(fontsize = 8),
151		name = NULL,
152		gp = grid::gpar(),
153		vp = NULL) {
154		# External margins need to be taken into account when defining the width of titles and footers
155		# because the text is split in advance depending on only the width of the viewport.
156	9x	if (any(as.numeric(outer_margins) > 0)) {
157	9x	width_titles <- width_titles - outer_margins[4]
158	9x	width_footnotes <- width_footnotes - outer_margins[4]
159		}
160
161	9x	st_titles <- split_text_grob(
162	9x	titles,
163	9x	x = 0, y = 1,
164	9x	just = c("left", "top"),
165	9x	width = width_titles,
166	9x	vp = grid::viewport(layout.pos.row = 1, layout.pos.col = 1),
167	9x	gp = gp_titles
168		)
169
170	9x	st_footnotes <- split_text_grob(
171	9x	footnotes,
172	9x	x = 0, y = 1,
173	9x	just = c("left", "top"),
174	9x	width = width_footnotes,
175	9x	vp = grid::viewport(layout.pos.row = 3, layout.pos.col = 1),
176	9x	gp = gp_footnotes
177		)
178
179	9x	pg_footnote <- grid::textGrob(
180	9x	paste("\n", page),
181	9x	x = 1, y = 0,
182	9x	just = c("right", "bottom"),
183	9x	vp = grid::viewport(layout.pos.row = 4, layout.pos.col = 1),
184	9x	gp = gp_footnotes
185		)
186
187		# Initial decoration of the grob -> border, paddings, and margins are used here
188	9x	main_plot <- grid::gTree(
189	9x	children = grid::gList(
190	9x	if (border) grid::rectGrob(),
191	9x	grid::gTree(
192	9x	children = grid::gList(
193	9x	grob
194		),
195	9x	vp = grid::plotViewport(margins = padding) # innermost margins of the grob plot
196		)
197		),
198	9x	vp = grid::vpStack(
199	9x	grid::viewport(layout.pos.row = 2, layout.pos.col = 1),
200	9x	grid::plotViewport(margins = margins) # margins around the border plot
201		)
202		)
203
204	9x	grid::gTree(
205	9x	grob = grob,
206	9x	titles = titles,
207	9x	footnotes = footnotes,
208	9x	page = page,
209	9x	width_titles = width_titles,
210	9x	width_footnotes = width_footnotes,
211	9x	outer_margins = outer_margins,
212	9x	gp_titles = gp_titles,
213	9x	gp_footnotes = gp_footnotes,
214	9x	children = grid::gList(
215	9x	grid::gTree(
216	9x	children = grid::gList(
217	9x	st_titles,
218	9x	main_plot, # main plot with border, padding, and margins
219	9x	st_footnotes,
220	9x	pg_footnote
221		),
222	9x	childrenvp = NULL,
223	9x	name = "titles_grob_footnotes",
224	9x	vp = grid::vpStack(
225	9x	grid::plotViewport(margins = outer_margins), # Main external margins
226	9x	grid::viewport(
227	9x	layout = grid::grid.layout(
228	9x	nrow = 4, ncol = 1,
229	9x	heights = grid::unit.c(
230	9x	grid::grobHeight(st_titles),
231	9x	grid::unit(1, "null"),
232	9x	grid::grobHeight(st_footnotes),
233	9x	grid::grobHeight(pg_footnote)
234		)
235		)
236		)
237		)
238		)
239		),
240	9x	name = name,
241	9x	gp = gp,
242	9x	vp = vp,
243	9x	cl = "decoratedGrob"
244		)
245		}
246
247		# nocov start
248		#' @importFrom grid validDetails
249		#' @noRd
250		validDetails.decoratedGrob <- function(x) {
251		checkmate::assert_character(x$titles)
252		checkmate::assert_character(x$footnotes)
253
254		if (!is.null(x$grob)) {
255		checkmate::assert_true(grid::is.grob(x$grob))
256		}
257		if (length(x$page) == 1) {
258		checkmate::assert_character(x$page)
259		}
260		if (!grid::is.unit(x$outer_margins)) {
261		checkmate::assert_vector(x$outer_margins, len = 4)
262		}
263		if (!grid::is.unit(x$margins)) {
264		checkmate::assert_vector(x$margins, len = 4)
265		}
266		if (!grid::is.unit(x$padding)) {
267		checkmate::assert_vector(x$padding, len = 4)
268		}
269
270		x
271		}
272
273		#' @importFrom grid widthDetails
274		#' @noRd
275		widthDetails.decoratedGrob <- function(x) {
276		grid::unit(1, "null")
277		}
278
279		#' @importFrom grid heightDetails
280		#' @noRd
281		heightDetails.decoratedGrob <- function(x) {
282		grid::unit(1, "null")
283		}
284
285		#' Split text according to available text width
286		#'
287		#' Dynamically wrap text.
288		#'
289		#' @inheritParams grid::grid.text
290		#' @param text (`string`)\cr the text to wrap.
291		#' @param width (`grid::unit`)\cr a unit object specifying maximum width of text.
292		#'
293		#' @return A text `grob`.
294		#'
295		#' @details This code is taken from `R Graphics by Paul Murell, 2nd edition`
296		#'
297		#' @keywords internal
298		split_text_grob <- function(text,
299		x = grid::unit(0.5, "npc"),
300		y = grid::unit(0.5, "npc"),
301		width = grid::unit(1, "npc"),
302		just = "centre",
303		hjust = NULL,
304		vjust = NULL,
305		default.units = "npc", # nolint
306		name = NULL,
307		gp = grid::gpar(),
308		vp = NULL) {
309		text <- gsub("\\\\n", "\n", text) # fixing cases of mixed behavior (\n and \\n)
310
311		if (!grid::is.unit(x)) x <- grid::unit(x, default.units)
312		if (!grid::is.unit(y)) y <- grid::unit(y, default.units)
313		if (!grid::is.unit(width)) width <- grid::unit(width, default.units)
314		if (grid::unitType(x) %in% c("sum", "min", "max")) x <- grid::convertUnit(x, default.units)
315		if (grid::unitType(y) %in% c("sum", "min", "max")) y <- grid::convertUnit(y, default.units)
316		if (grid::unitType(width) %in% c("sum", "min", "max")) width <- grid::convertUnit(width, default.units)
317
318		if (length(gp) > 0) { # account for effect of gp on text width -> it was bugging when text was empty
319		horizontal_npc_width_no_gp <- grid::convertWidth(
320		grid::grobWidth(
321		grid::textGrob(
322		paste0(text, collapse = "\n")
323		)
324		), "npc",
325		valueOnly = TRUE
326		)
327		horizontal_npc_width_with_gp <- grid::convertWidth(grid::grobWidth(
328		grid::textGrob(
329		paste0(text, collapse = "\n"),
330		gp = gp
331		)
332		), "npc", valueOnly = TRUE)
333
334		# Adapting width to the input gpar (it is normalized so does not matter what is text)
335		width <- width * horizontal_npc_width_no_gp / horizontal_npc_width_with_gp
336		}
337
338		## if it is a fixed unit then we do not need to recalculate when viewport resized
339		if (!inherits(width, "unit.arithmetic") && !is.null(attr(width, "unit")) &&
340		attr(width, "unit") %in% c("cm", "inches", "mm", "points", "picas", "bigpts", "dida", "cicero", "scaledpts")) { # nolint
341		attr(text, "fixed_text") <- paste(vapply(text, split_string, character(1), width = width), collapse = "\n")
342		}
343
344		# Fix for split_string in case of residual \n (otherwise is counted as character)
345		text2 <- unlist(
346		strsplit(
347		paste0(text, collapse = "\n"), # for "" cases
348		"\n"
349		)
350		)
351
352		# Final grid text with cat-friendly split_string
353		grid::grid.text(
354		label = split_string(text2, width),
355		x = x, y = y,
356		just = just,
357		hjust = hjust,
358		vjust = vjust,
359		rot = 0,
360		check.overlap = FALSE,
361		name = name,
362		gp = gp,
363		vp = vp,
364		draw = FALSE
365		)
366		}
367
368		#' @importFrom grid validDetails
369		#' @noRd
370		validDetails.dynamicSplitText <- function(x) {
371		checkmate::assert_character(x$text)
372		checkmate::assert_true(grid::is.unit(x$width))
373		checkmate::assert_vector(x$width, len = 1)
374		x
375		}
376
377		#' @importFrom grid heightDetails
378		#' @noRd
379		heightDetails.dynamicSplitText <- function(x) {
380		txt <- if (!is.null(attr(x$text, "fixed_text"))) {
381		attr(x$text, "fixed_text")
382		} else {
383		paste(vapply(x$text, split_string, character(1), width = x$width), collapse = "\n")
384		}
385		grid::stringHeight(txt)
386		}
387
388		#' @importFrom grid widthDetails
389		#' @noRd
390		widthDetails.dynamicSplitText <- function(x) {
391		x$width
392		}
393
394		#' @importFrom grid drawDetails
395		#' @noRd
396		drawDetails.dynamicSplitText <- function(x, recording) {
397		txt <- if (!is.null(attr(x$text, "fixed_text"))) {
398		attr(x$text, "fixed_text")
399		} else {
400		paste(vapply(x$text, split_string, character(1), width = x$width), collapse = "\n")
401		}
402
403		x$width <- NULL
404		x$label <- txt
405		x$text <- NULL
406		class(x) <- c("text", class(x)[-1])
407
408		grid::grid.draw(x)
409		}
410		# nocov end
411
412		# Adapted from Paul Murell R Graphics 2nd Edition
413		# https://www.stat.auckland.ac.nz/~paul/RG2e/interactgrid-splittext.R
414		split_string <- function(text, width) {
415	26x	strings <- strsplit(text, " ")
416	26x	out_string <- NA
417	26x	for (string_i in seq_along(strings)) {
418	48x	newline_str <- strings[[string_i]]
419	6x	if (length(newline_str) == 0) newline_str <- ""
420	48x	if (is.na(out_string[string_i])) {
421	48x	out_string[string_i] <- newline_str[[1]][[1]]
422	48x	linewidth <- grid::stringWidth(out_string[string_i])
423		}
424	48x	gapwidth <- grid::stringWidth(" ")
425	48x	availwidth <- as.numeric(width)
426	48x	if (length(newline_str) > 1) {
427	12x	for (i in seq(2, length(newline_str))) {
428	184x	width_i <- grid::stringWidth(newline_str[i])
429		# Main conversion of allowed text width -> npc units are 0<npc<1. External viewport is used for conversion
430	184x	if (grid::convertWidth(linewidth + gapwidth + width_i, grid::unitType(width), valueOnly = TRUE) < availwidth) {
431	177x	sep <- " "
432	177x	linewidth <- linewidth + gapwidth + width_i
433		} else {
434	7x	sep <- "\n"
435	7x	linewidth <- width_i
436		}
437	184x	out_string[string_i] <- paste(out_string[string_i], newline_str[i], sep = sep)
438		}
439		}
440		}
441	26x	paste(out_string, collapse = "\n")
442		}
443
444		#' Update page number
445		#'
446		#' Automatically updates page number.
447		#'
448		#' @param npages (`numeric(1)`)\cr total number of pages.
449		#' @param ... arguments passed on to [decorate_grob()].
450		#'
451		#' @return Closure that increments the page number.
452		#'
453		#' @keywords internal
454		decorate_grob_factory <- function(npages, ...) {
455	2x	current_page <- 0
456	2x	function(grob) {
457	7x	current_page <<- current_page + 1
458	7x	if (current_page > npages) {
459	1x	stop(paste("current page is", current_page, "but max.", npages, "specified."))
460		}
461	6x	decorate_grob(grob = grob, page = paste("Page", current_page, "of", npages), ...)
462		}
463		}
464
465		#' Decorate set of `grob`s and add page numbering
466		#'
467		#' @description `r lifecycle::badge("stable")`
468		#'
469		#' Note that this uses the [decorate_grob_factory()] function.
470		#'
471		#' @param grobs (`list` of `grob`)\cr a list of grid grobs.
472		#' @param ... arguments passed on to [decorate_grob()].
473		#'
474		#' @return A decorated grob.
475		#'
476		#' @examples
477		#' library(ggplot2)
478		#' library(grid)
479		#' g <- with(data = iris, {
480		#' list(
481		#' ggplot2::ggplotGrob(
482		#' ggplot2::ggplot(mapping = aes(Sepal.Length, Sepal.Width, col = Species)) +
483		#' ggplot2::geom_point()
484		#' ),
485		#' ggplot2::ggplotGrob(
486		#' ggplot2::ggplot(mapping = aes(Sepal.Length, Petal.Length, col = Species)) +
487		#' ggplot2::geom_point()
488		#' ),
489		#' ggplot2::ggplotGrob(
490		#' ggplot2::ggplot(mapping = aes(Sepal.Length, Petal.Width, col = Species)) +
491		#' ggplot2::geom_point()
492		#' ),
493		#' ggplot2::ggplotGrob(
494		#' ggplot2::ggplot(mapping = aes(Sepal.Width, Petal.Length, col = Species)) +
495		#' ggplot2::geom_point()
496		#' ),
497		#' ggplot2::ggplotGrob(
498		#' ggplot2::ggplot(mapping = aes(Sepal.Width, Petal.Width, col = Species)) +
499		#' ggplot2::geom_point()
500		#' ),
501		#' ggplot2::ggplotGrob(
502		#' ggplot2::ggplot(mapping = aes(Petal.Length, Petal.Width, col = Species)) +
503		#' ggplot2::geom_point()
504		#' )
505		#' )
506		#' })
507		#' lg <- decorate_grob_set(grobs = g, titles = "Hello\nOne\nTwo\nThree", footnotes = "")
508		#'
509		#' draw_grob(lg[[1]])
510		#' draw_grob(lg[[2]])
511		#' draw_grob(lg[[6]])
512		#'
513		#' @export
514		decorate_grob_set <- function(grobs, ...) {
515	1x	n <- length(grobs)
516	1x	lgf <- decorate_grob_factory(npages = n, ...)
517	1x	lapply(grobs, lgf)
518		}

1		#' Survival time point analysis
2		#'
3		#' @description `r lifecycle::badge("stable")`
4		#'
5		#' The analyze function [surv_timepoint()] creates a layout element to analyze patient survival rates and difference
6		#' of survival rates between groups at a given time point. The primary analysis variable `vars` is the time variable.
7		#' Other required inputs are `time_point`, the numeric time point of interest, and `is_event`, a variable that
8		#' indicates whether or not an event has occurred. The `method` argument is used to specify whether you want to analyze
9		#' survival estimations (`"surv"`), difference in survival with the control (`"surv_diff"`), or both of these
10		#' (`"both"`).
11		#'
12		#' @inheritParams argument_convention
13		#' @inheritParams s_surv_time
14		#' @param time_point (`numeric(1)`)\cr survival time point of interest.
15		#' @param control (`list`)\cr parameters for comparison details, specified by using the helper function
16		#' [control_surv_timepoint()]. Some possible parameter options are:
17		#' * `conf_level` (`proportion`)\cr confidence level of the interval for survival rate.
18		#' * `conf_type` (`string`)\cr confidence interval type. Options are "plain" (default), "log", "log-log",
19		#' see more in [survival::survfit()]. Note option "none" is no longer supported.
20		#' @param method (`string`)\cr `"surv"` (survival estimations), `"surv_diff"` (difference in survival with the
21		#' control), or `"both"`.
22		#' @param table_names_suffix (`string`)\cr optional suffix for the `table_names` used for the `rtables` to
23		#' avoid warnings from duplicate table names.
24		#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("surv_timepoint")`
25		#' to see available statistics for this function.
26		#' @param .indent_mods (named `integer`)\cr indent modifiers for the labels. Each element of the vector
27		#' should be a name-value pair with name corresponding to a statistic specified in `.stats` and value the indentation
28		#' for that statistic's row label.
29		#'
30		#' @name survival_timepoint
31		#' @order 1
32		NULL
33
34		#' @describeIn survival_timepoint Statistics function which analyzes survival rate.
35		#'
36		#' @return
37		#' * `s_surv_timepoint()` returns the statistics:
38		#' * `pt_at_risk`: Patients remaining at risk.
39		#' * `event_free_rate`: Event-free rate (%).
40		#' * `rate_se`: Standard error of event free rate.
41		#' * `rate_ci`: Confidence interval for event free rate.
42		#'
43		#' @keywords internal
44		s_surv_timepoint <- function(df,
45		.var,
46		time_point,
47		is_event,
48		control = control_surv_timepoint()) {
49	23x	checkmate::assert_string(.var)
50	23x	assert_df_with_variables(df, list(tte = .var, is_event = is_event))
51	23x	checkmate::assert_numeric(df[[.var]], min.len = 1, any.missing = FALSE)
52	23x	checkmate::assert_number(time_point)
53	23x	checkmate::assert_logical(df[[is_event]], min.len = 1, any.missing = FALSE)
54
55	23x	conf_type <- control$conf_type
56	23x	conf_level <- control$conf_level
57
58	23x	formula <- stats::as.formula(paste0("survival::Surv(", .var, ", ", is_event, ") ~ 1"))
59	23x	srv_fit <- survival::survfit(
60	23x	formula = formula,
61	23x	data = df,
62	23x	conf.int = conf_level,
63	23x	conf.type = conf_type
64		)
65	23x	s_srv_fit <- summary(srv_fit, times = time_point, extend = TRUE)
66	23x	df_srv_fit <- as.data.frame(s_srv_fit[c("time", "n.risk", "surv", "lower", "upper", "std.err")])
67	23x	if (df_srv_fit[["n.risk"]] == 0) {
68	1x	pt_at_risk <- event_free_rate <- rate_se <- NA_real_
69	1x	rate_ci <- c(NA_real_, NA_real_)
70		} else {
71	22x	pt_at_risk <- df_srv_fit$n.risk
72	22x	event_free_rate <- df_srv_fit$surv
73	22x	rate_se <- df_srv_fit$std.err
74	22x	rate_ci <- c(df_srv_fit$lower, df_srv_fit$upper)
75		}
76	23x	list(
77	23x	pt_at_risk = formatters::with_label(pt_at_risk, "Patients remaining at risk"),
78	23x	event_free_rate = formatters::with_label(event_free_rate * 100, "Event Free Rate (%)"),
79	23x	rate_se = formatters::with_label(rate_se * 100, "Standard Error of Event Free Rate"),
80	23x	rate_ci = formatters::with_label(rate_ci * 100, f_conf_level(conf_level))
81		)
82		}
83
84		#' @describeIn survival_timepoint Formatted analysis function which is used as `afun` in `surv_timepoint()`
85		#' when `method = "surv"`.
86		#'
87		#' @return
88		#' * `a_surv_timepoint()` returns the corresponding list with formatted [rtables::CellValue()].
89		#'
90		#' @keywords internal
91		a_surv_timepoint <- make_afun(
92		s_surv_timepoint,
93		.indent_mods = c(
94		pt_at_risk = 0L,
95		event_free_rate = 0L,
96		rate_se = 1L,
97		rate_ci = 1L
98		),
99		.formats = c(
100		pt_at_risk = "xx",
101		event_free_rate = "xx.xx",
102		rate_se = "xx.xx",
103		rate_ci = "(xx.xx, xx.xx)"
104		)
105		)
106
107		#' @describeIn survival_timepoint Statistics function which analyzes difference between two survival rates.
108		#'
109		#' @return
110		#' * `s_surv_timepoint_diff()` returns the statistics:
111		#' * `rate_diff`: Event-free rate difference between two groups.
112		#' * `rate_diff_ci`: Confidence interval for the difference.
113		#' * `ztest_pval`: p-value to test the difference is 0.
114		#'
115		#' @keywords internal
116		s_surv_timepoint_diff <- function(df,
117		.var,
118		.ref_group,
119		.in_ref_col,
120		time_point,
121		control = control_surv_timepoint(),
122		...) {
123	2x	if (.in_ref_col) {
124	!	return(
125	!	list(
126	!	rate_diff = formatters::with_label("", "Difference in Event Free Rate"),
127	!	rate_diff_ci = formatters::with_label("", f_conf_level(control$conf_level)),
128	!	ztest_pval = formatters::with_label("", "p-value (Z-test)")
129		)
130		)
131		}
132	2x	data <- rbind(.ref_group, df)
133	2x	group <- factor(rep(c("ref", "x"), c(nrow(.ref_group), nrow(df))), levels = c("ref", "x"))
134	2x	res_per_group <- lapply(split(data, group), function(x) {
135	4x	s_surv_timepoint(df = x, .var = .var, time_point = time_point, control = control, ...)
136		})
137
138	2x	res_x <- res_per_group[[2]]
139	2x	res_ref <- res_per_group[[1]]
140	2x	rate_diff <- res_x$event_free_rate - res_ref$event_free_rate
141	2x	se_diff <- sqrt(res_x$rate_se^2 + res_ref$rate_se^2)
142
143	2x	qs <- c(-1, 1) * stats::qnorm(1 - (1 - control$conf_level) / 2)
144	2x	rate_diff_ci <- rate_diff + qs * se_diff
145	2x	ztest_pval <- if (is.na(rate_diff)) {
146	2x	NA
147		} else {
148	2x	2 * (1 - stats::pnorm(abs(rate_diff) / se_diff))
149		}
150	2x	list(
151	2x	rate_diff = formatters::with_label(rate_diff, "Difference in Event Free Rate"),
152	2x	rate_diff_ci = formatters::with_label(rate_diff_ci, f_conf_level(control$conf_level)),
153	2x	ztest_pval = formatters::with_label(ztest_pval, "p-value (Z-test)")
154		)
155		}
156
157		#' @describeIn survival_timepoint Formatted analysis function which is used as `afun` in `surv_timepoint()`
158		#' when `method = "surv_diff"`.
159		#'
160		#' @return
161		#' * `a_surv_timepoint_diff()` returns the corresponding list with formatted [rtables::CellValue()].
162		#'
163		#' @keywords internal
164		a_surv_timepoint_diff <- make_afun(
165		s_surv_timepoint_diff,
166		.formats = c(
167		rate_diff = "xx.xx",
168		rate_diff_ci = "(xx.xx, xx.xx)",
169		ztest_pval = "x.xxxx \| (<0.0001)"
170		)
171		)
172
173		#' @describeIn survival_timepoint Layout-creating function which can take statistics function arguments
174		#' and additional format arguments. This function is a wrapper for [rtables::analyze()].
175		#'
176		#' @return
177		#' * `surv_timepoint()` returns a layout object suitable for passing to further layouting functions,
178		#' or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
179		#' the statistics from `s_surv_timepoint()` and/or `s_surv_timepoint_diff()` to the table layout depending on
180		#' the value of `method`.
181		#'
182		#' @examples
183		#' library(dplyr)
184		#'
185		#' adtte_f <- tern_ex_adtte %>%
186		#' filter(PARAMCD == "OS") %>%
187		#' mutate(
188		#' AVAL = day2month(AVAL),
189		#' is_event = CNSR == 0
190		#' )
191		#'
192		#' # Survival at given time points.
193		#' basic_table() %>%
194		#' split_cols_by(var = "ARMCD", ref_group = "ARM A") %>%
195		#' add_colcounts() %>%
196		#' surv_timepoint(
197		#' vars = "AVAL",
198		#' var_labels = "Months",
199		#' is_event = "is_event",
200		#' time_point = 7
201		#' ) %>%
202		#' build_table(df = adtte_f)
203		#'
204		#' # Difference in survival at given time points.
205		#' basic_table() %>%
206		#' split_cols_by(var = "ARMCD", ref_group = "ARM A") %>%
207		#' add_colcounts() %>%
208		#' surv_timepoint(
209		#' vars = "AVAL",
210		#' var_labels = "Months",
211		#' is_event = "is_event",
212		#' time_point = 9,
213		#' method = "surv_diff",
214		#' .indent_mods = c("rate_diff" = 0L, "rate_diff_ci" = 2L, "ztest_pval" = 2L)
215		#' ) %>%
216		#' build_table(df = adtte_f)
217		#'
218		#' # Survival and difference in survival at given time points.
219		#' basic_table() %>%
220		#' split_cols_by(var = "ARMCD", ref_group = "ARM A") %>%
221		#' add_colcounts() %>%
222		#' surv_timepoint(
223		#' vars = "AVAL",
224		#' var_labels = "Months",
225		#' is_event = "is_event",
226		#' time_point = 9,
227		#' method = "both"
228		#' ) %>%
229		#' build_table(df = adtte_f)
230		#'
231		#' @export
232		#' @order 2
233		surv_timepoint <- function(lyt,
234		vars,
235		time_point,
236		is_event,
237		control = control_surv_timepoint(),
238		method = c("surv", "surv_diff", "both"),
239		na_str = default_na_str(),
240		nested = TRUE,
241		...,
242		table_names_suffix = "",
243		var_labels = "Time",
244		show_labels = "visible",
245		.stats = c(
246		"pt_at_risk", "event_free_rate", "rate_ci",
247		"rate_diff", "rate_diff_ci", "ztest_pval"
248		),
249		.formats = NULL,
250		.labels = NULL,
251		.indent_mods = if (method == "both") {
252	2x	c(rate_diff = 1L, rate_diff_ci = 2L, ztest_pval = 2L)
253		} else {
254	4x	c(rate_diff_ci = 1L, ztest_pval = 1L)
255		}) {
256	6x	method <- match.arg(method)
257	6x	checkmate::assert_string(table_names_suffix)
258
259	6x	extra_args <- list(time_point = time_point, is_event = is_event, control = control, ...)
260
261	6x	f <- list(
262	6x	surv = c("pt_at_risk", "event_free_rate", "rate_se", "rate_ci"),
263	6x	surv_diff = c("rate_diff", "rate_diff_ci", "ztest_pval")
264		)
265	6x	.stats <- h_split_param(.stats, .stats, f = f)
266	6x	.formats <- h_split_param(.formats, names(.formats), f = f)
267	6x	.labels <- h_split_param(.labels, names(.labels), f = f)
268	6x	.indent_mods <- h_split_param(.indent_mods, names(.indent_mods), f = f)
269
270	6x	afun_surv <- make_afun(
271	6x	a_surv_timepoint,
272	6x	.stats = .stats$surv,
273	6x	.formats = .formats$surv,
274	6x	.labels = .labels$surv,
275	6x	.indent_mods = .indent_mods$surv
276		)
277
278	6x	afun_surv_diff <- make_afun(
279	6x	a_surv_timepoint_diff,
280	6x	.stats = .stats$surv_diff,
281	6x	.formats = .formats$surv_diff,
282	6x	.labels = .labels$surv_diff,
283	6x	.indent_mods = .indent_mods$surv_diff
284		)
285
286	6x	time_point <- extra_args$time_point
287
288	6x	for (i in seq_along(time_point)) {
289	6x	extra_args[["time_point"]] <- time_point[i]
290
291	6x	if (method %in% c("surv", "both")) {
292	4x	lyt <- analyze(
293	4x	lyt,
294	4x	vars,
295	4x	var_labels = paste(time_point[i], var_labels),
296	4x	table_names = paste0("surv_", time_point[i], table_names_suffix),
297	4x	show_labels = show_labels,
298	4x	afun = afun_surv,
299	4x	na_str = na_str,
300	4x	nested = nested,
301	4x	extra_args = extra_args
302		)
303		}
304
305	6x	if (method %in% c("surv_diff", "both")) {
306	4x	lyt <- analyze(
307	4x	lyt,
308	4x	vars,
309	4x	var_labels = paste(time_point[i], var_labels),
310	4x	table_names = paste0("surv_diff_", time_point[i], table_names_suffix),
311	4x	show_labels = ifelse(method == "both", "hidden", show_labels),
312	4x	afun = afun_surv_diff,
313	4x	na_str = na_str,
314	4x	nested = nested,
315	4x	extra_args = extra_args
316		)
317		}
318		}
319	6x	lyt
320		}

1		#' Create a STEP graph
2		#'
3		#' @description `r lifecycle::badge("stable")`
4		#'
5		#' Based on the STEP results, creates a `ggplot` graph showing the estimated HR or OR
6		#' along the continuous biomarker value subgroups.
7		#'
8		#' @param df (`tibble`)\cr result of [tidy.step()].
9		#' @param use_percentile (`flag`)\cr whether to use percentiles for the x axis or actual
10		#' biomarker values.
11		#' @param est (named `list`)\cr `col` and `lty` settings for estimate line.
12		#' @param ci_ribbon (named `list` or `NULL`)\cr `fill` and `alpha` settings for the confidence interval
13		#' ribbon area, or `NULL` to not plot a CI ribbon.
14		#' @param col (`character`)\cr color(s).
15		#'
16		#' @return A `ggplot` STEP graph.
17		#'
18		#' @seealso Custom tidy method [tidy.step()].
19		#'
20		#' @examples
21		#' library(nestcolor)
22		#' library(survival)
23		#' lung$sex <- factor(lung$sex)
24		#'
25		#' # Survival example.
26		#' vars <- list(
27		#' time = "time",
28		#' event = "status",
29		#' arm = "sex",
30		#' biomarker = "age"
31		#' )
32		#'
33		#' step_matrix <- fit_survival_step(
34		#' variables = vars,
35		#' data = lung,
36		#' control = c(control_coxph(), control_step(num_points = 10, degree = 2))
37		#' )
38		#' step_data <- broom::tidy(step_matrix)
39		#'
40		#' # Default plot.
41		#' g_step(step_data)
42		#'
43		#' # Add the reference 1 horizontal line.
44		#' library(ggplot2)
45		#' g_step(step_data) +
46		#' ggplot2::geom_hline(ggplot2::aes(yintercept = 1), linetype = 2)
47		#'
48		#' # Use actual values instead of percentiles, different color for estimate and no CI,
49		#' # use log scale for y axis.
50		#' g_step(
51		#' step_data,
52		#' use_percentile = FALSE,
53		#' est = list(col = "blue", lty = 1),
54		#' ci_ribbon = NULL
55		#' ) + scale_y_log10()
56		#'
57		#' # Adding another curve based on additional column.
58		#' step_data$extra <- exp(step_data$`Percentile Center`)
59		#' g_step(step_data) +
60		#' ggplot2::geom_line(ggplot2::aes(y = extra), linetype = 2, color = "green")
61		#'
62		#' # Response example.
63		#' vars <- list(
64		#' response = "status",
65		#' arm = "sex",
66		#' biomarker = "age"
67		#' )
68		#'
69		#' step_matrix <- fit_rsp_step(
70		#' variables = vars,
71		#' data = lung,
72		#' control = c(
73		#' control_logistic(response_definition = "I(response == 2)"),
74		#' control_step()
75		#' )
76		#' )
77		#' step_data <- broom::tidy(step_matrix)
78		#' g_step(step_data)
79		#'
80		#' @export
81		g_step <- function(df,
82		use_percentile = "Percentile Center" %in% names(df),
83		est = list(col = "blue", lty = 1),
84		ci_ribbon = list(fill = getOption("ggplot2.discrete.colour")[1], alpha = 0.5),
85		col = getOption("ggplot2.discrete.colour")) {
86	2x	checkmate::assert_tibble(df)
87	2x	checkmate::assert_flag(use_percentile)
88	2x	checkmate::assert_character(col, null.ok = TRUE)
89	2x	checkmate::assert_list(est, names = "named")
90	2x	checkmate::assert_list(ci_ribbon, names = "named", null.ok = TRUE)
91
92	2x	x_var <- ifelse(use_percentile, "Percentile Center", "Interval Center")
93	2x	df$x <- df[[x_var]]
94	2x	attrs <- attributes(df)
95	2x	df$y <- df[[attrs$estimate]]
96
97		# Set legend names. To be modified also at call level
98	2x	legend_names <- c("Estimate", "CI 95%")
99
100	2x	p <- ggplot2::ggplot(df, ggplot2::aes(x = .data[["x"]], y = .data[["y"]]))
101
102	2x	if (!is.null(col)) {
103	2x	p <- p +
104	2x	ggplot2::scale_color_manual(values = col)
105		}
106
107	2x	if (!is.null(ci_ribbon)) {
108	1x	if (is.null(ci_ribbon$fill)) {
109	!	ci_ribbon$fill <- "lightblue"
110		}
111	1x	p <- p + ggplot2::geom_ribbon(
112	1x	ggplot2::aes(
113	1x	ymin = .data[["ci_lower"]], ymax = .data[["ci_upper"]],
114	1x	fill = legend_names[2]
115		),
116	1x	alpha = ci_ribbon$alpha
117		) +
118	1x	scale_fill_manual(
119	1x	name = "", values = c("CI 95%" = ci_ribbon$fill)
120		)
121		}
122	2x	suppressMessages(p <- p +
123	2x	ggplot2::geom_line(
124	2x	ggplot2::aes(y = .data[["y"]], color = legend_names[1]),
125	2x	linetype = est$lty
126		) +
127	2x	scale_colour_manual(
128	2x	name = "", values = c("Estimate" = "blue")
129		))
130
131	2x	p <- p + ggplot2::labs(x = attrs$biomarker, y = attrs$estimate)
132	2x	if (use_percentile) {
133	1x	p <- p + ggplot2::scale_x_continuous(labels = scales::percent)
134		}
135	2x	p
136		}
137
138		#' Custom tidy method for STEP results
139		#'
140		#' @description `r lifecycle::badge("stable")`
141		#'
142		#' Tidy the STEP results into a `tibble` format ready for plotting.
143		#'
144		#' @param x (`matrix`)\cr results from [fit_survival_step()].
145		#' @param ... not used.
146		#'
147		#' @return A `tibble` with one row per STEP subgroup. The estimates and CIs are on the HR or OR scale,
148		#' respectively. Additional attributes carry metadata also used for plotting.
149		#'
150		#' @seealso [g_step()] which consumes the result from this function.
151		#'
152		#' @method tidy step
153		#'
154		#' @examples
155		#' library(survival)
156		#' lung$sex <- factor(lung$sex)
157		#' vars <- list(
158		#' time = "time",
159		#' event = "status",
160		#' arm = "sex",
161		#' biomarker = "age"
162		#' )
163		#' step_matrix <- fit_survival_step(
164		#' variables = vars,
165		#' data = lung,
166		#' control = c(control_coxph(), control_step(num_points = 10, degree = 2))
167		#' )
168		#' broom::tidy(step_matrix)
169		#'
170		#' @export
171		tidy.step <- function(x, ...) { # nolint
172	7x	checkmate::assert_class(x, "step")
173	7x	dat <- as.data.frame(x)
174	7x	nams <- names(dat)
175	7x	is_surv <- "loghr" %in% names(dat)
176	7x	est_var <- ifelse(is_surv, "loghr", "logor")
177	7x	new_est_var <- ifelse(is_surv, "Hazard Ratio", "Odds Ratio")
178	7x	new_y_vars <- c(new_est_var, c("ci_lower", "ci_upper"))
179	7x	names(dat)[match(est_var, nams)] <- new_est_var
180	7x	dat[, new_y_vars] <- exp(dat[, new_y_vars])
181	7x	any_is_na <- any(is.na(dat[, new_y_vars]))
182	7x	any_is_very_large <- any(abs(dat[, new_y_vars]) > 1e10, na.rm = TRUE)
183	7x	if (any_is_na) {
184	2x	warning(paste(
185	2x	"Missing values in the point estimate or CI columns,",
186	2x	"this will lead to holes in the `g_step()` plot"
187		))
188		}
189	7x	if (any_is_very_large) {
190	2x	warning(paste(
191	2x	"Very large absolute values in the point estimate or CI columns,",
192	2x	"consider adding `scale_y_log10()` to the `g_step()` result for plotting"
193		))
194		}
195	7x	if (any_is_na \|\| any_is_very_large) {
196	4x	warning("Consider using larger `bandwidth`, less `num_points` in `control_step()` settings for fitting")
197		}
198	7x	structure(
199	7x	tibble::as_tibble(dat),
200	7x	estimate = new_est_var,
201	7x	biomarker = attr(x, "variables")$biomarker,
202	7x	ci = f_conf_level(attr(x, "control")$conf_level)
203		)
204		}

1		#' Control function for Cox regression
2		#'
3		#' @description `r lifecycle::badge("stable")`
4		#'
5		#' Sets a list of parameters for Cox regression fit. Used internally.
6		#'
7		#' @inheritParams argument_convention
8		#' @param pval_method (`string`)\cr the method used for estimation of p.values; `wald` (default) or `likelihood`.
9		#' @param interaction (`flag`)\cr if `TRUE`, the model includes the interaction between the studied
10		#' treatment and candidate covariate. Note that for univariate models without treatment arm, and
11		#' multivariate models, no interaction can be used so that this needs to be `FALSE`.
12		#' @param ties (`string`)\cr among `exact` (equivalent to `DISCRETE` in SAS), `efron` and `breslow`,
13		#' see [survival::coxph()]. Note: there is no equivalent of SAS `EXACT` method in R.
14		#'
15		#' @return A `list` of items with names corresponding to the arguments.
16		#'
17		#' @seealso [fit_coxreg_univar()] and [fit_coxreg_multivar()].
18		#'
19		#' @examples
20		#' control_coxreg()
21		#'
22		#' @export
23		control_coxreg <- function(pval_method = c("wald", "likelihood"),
24		ties = c("exact", "efron", "breslow"),
25		conf_level = 0.95,
26		interaction = FALSE) {
27	55x	pval_method <- match.arg(pval_method)
28	55x	ties <- match.arg(ties)
29	55x	checkmate::assert_flag(interaction)
30	55x	assert_proportion_value(conf_level)
31	55x	list(
32	55x	pval_method = pval_method,
33	55x	ties = ties,
34	55x	conf_level = conf_level,
35	55x	interaction = interaction
36		)
37		}
38
39		#' Custom tidy methods for Cox regression
40		#'
41		#' @description `r lifecycle::badge("stable")`
42		#'
43		#' @inheritParams argument_convention
44		#' @param x (`list`)\cr result of the Cox regression model fitted by [fit_coxreg_univar()] (for univariate models)
45		#' or [fit_coxreg_multivar()] (for multivariate models).
46		#'
47		#' @return [broom::tidy()] returns:
48		#' * For `summary.coxph` objects, a `data.frame` with columns: `Pr(>\|z\|)`, `exp(coef)`, `exp(-coef)`, `lower .95`,
49		#' `upper .95`, `level`, and `n`.
50		#' * For `coxreg.univar` objects, a `data.frame` with columns: `effect`, `term`, `term_label`, `level`, `n`, `hr`,
51		#' `lcl`, `ucl`, `pval`, and `ci`.
52		#' * For `coxreg.multivar` objects, a `data.frame` with columns: `term`, `pval`, `term_label`, `hr`, `lcl`, `ucl`,
53		#' `level`, and `ci`.
54		#'
55		#' @seealso [cox_regression]
56		#'
57		#' @name tidy_coxreg
58		NULL
59
60		#' @describeIn tidy_coxreg Custom tidy method for [survival::coxph()] summary results.
61		#'
62		#' Tidy the [survival::coxph()] results into a `data.frame` to extract model results.
63		#'
64		#' @method tidy summary.coxph
65		#'
66		#' @examples
67		#' library(survival)
68		#' library(broom)
69		#'
70		#' set.seed(1, kind = "Mersenne-Twister")
71		#'
72		#' dta_bladder <- with(
73		#' data = bladder[bladder$enum < 5, ],
74		#' data.frame(
75		#' time = stop,
76		#' status = event,
77		#' armcd = as.factor(rx),
78		#' covar1 = as.factor(enum),
79		#' covar2 = factor(
80		#' sample(as.factor(enum)),
81		#' levels = 1:4, labels = c("F", "F", "M", "M")
82		#' )
83		#' )
84		#' )
85		#' labels <- c("armcd" = "ARM", "covar1" = "A Covariate Label", "covar2" = "Sex (F/M)")
86		#' formatters::var_labels(dta_bladder)[names(labels)] <- labels
87		#' dta_bladder$age <- sample(20:60, size = nrow(dta_bladder), replace = TRUE)
88		#'
89		#' formula <- "survival::Surv(time, status) ~ armcd + covar1"
90		#' msum <- summary(coxph(stats::as.formula(formula), data = dta_bladder))
91		#' tidy(msum)
92		#'
93		#' @export
94		tidy.summary.coxph <- function(x, # nolint
95		...) {
96	199x	checkmate::assert_class(x, "summary.coxph")
97	199x	pval <- x$coefficients
98	199x	confint <- x$conf.int
99	199x	levels <- rownames(pval)
100
101	199x	pval <- tibble::as_tibble(pval)
102	199x	confint <- tibble::as_tibble(confint)
103
104	199x	ret <- cbind(pval[, grepl("Pr", names(pval))], confint)
105	199x	ret$level <- levels
106	199x	ret$n <- x[["n"]]
107	199x	ret
108		}
109
110		#' @describeIn tidy_coxreg Custom tidy method for a univariate Cox regression.
111		#'
112		#' Tidy up the result of a Cox regression model fitted by [fit_coxreg_univar()].
113		#'
114		#' @method tidy coxreg.univar
115		#'
116		#' @examples
117		#' ## Cox regression: arm + 1 covariate.
118		#' mod1 <- fit_coxreg_univar(
119		#' variables = list(
120		#' time = "time", event = "status", arm = "armcd",
121		#' covariates = "covar1"
122		#' ),
123		#' data = dta_bladder,
124		#' control = control_coxreg(conf_level = 0.91)
125		#' )
126		#'
127		#' ## Cox regression: arm + 1 covariate + interaction, 2 candidate covariates.
128		#' mod2 <- fit_coxreg_univar(
129		#' variables = list(
130		#' time = "time", event = "status", arm = "armcd",
131		#' covariates = c("covar1", "covar2")
132		#' ),
133		#' data = dta_bladder,
134		#' control = control_coxreg(conf_level = 0.91, interaction = TRUE)
135		#' )
136		#'
137		#' tidy(mod1)
138		#' tidy(mod2)
139		#'
140		#' @export
141		tidy.coxreg.univar <- function(x, # nolint
142		...) {
143	38x	checkmate::assert_class(x, "coxreg.univar")
144	38x	mod <- x$mod
145	38x	vars <- c(x$vars$arm, x$vars$covariates)
146	38x	has_arm <- "arm" %in% names(x$vars)
147
148	38x	result <- if (!has_arm) {
149	5x	Map(
150	5x	mod = mod, vars = vars,
151	5x	f = function(mod, vars) {
152	6x	h_coxreg_multivar_extract(
153	6x	var = vars,
154	6x	data = x$data,
155	6x	mod = mod,
156	6x	control = x$control
157		)
158		}
159		)
160	38x	} else if (x$control$interaction) {
161	12x	Map(
162	12x	mod = mod, covar = vars,
163	12x	f = function(mod, covar) {
164	26x	h_coxreg_extract_interaction(
165	26x	effect = x$vars$arm, covar = covar, mod = mod, data = x$data,
166	26x	at = x$at, control = x$control
167		)
168		}
169		)
170		} else {
171	21x	Map(
172	21x	mod = mod, vars = vars,
173	21x	f = function(mod, vars) {
174	53x	h_coxreg_univar_extract(
175	53x	effect = x$vars$arm, covar = vars, data = x$data, mod = mod,
176	53x	control = x$control
177		)
178		}
179		)
180		}
181	38x	result <- do.call(rbind, result)
182
183	38x	result$ci <- Map(lcl = result$lcl, ucl = result$ucl, f = function(lcl, ucl) c(lcl, ucl))
184	38x	result$n <- lapply(result$n, empty_vector_if_na)
185	38x	result$ci <- lapply(result$ci, empty_vector_if_na)
186	38x	result$hr <- lapply(result$hr, empty_vector_if_na)
187	38x	if (x$control$interaction) {
188	12x	result$pval_inter <- lapply(result$pval_inter, empty_vector_if_na)
189		# Remove interaction p-values due to change in specifications.
190	12x	result$pval[result$effect != "Treatment:"] <- NA
191		}
192	38x	result$pval <- lapply(result$pval, empty_vector_if_na)
193	38x	attr(result, "conf_level") <- x$control$conf_level
194	38x	result
195		}
196
197		#' @describeIn tidy_coxreg Custom tidy method for a multivariate Cox regression.
198		#'
199		#' Tidy up the result of a Cox regression model fitted by [fit_coxreg_multivar()].
200		#'
201		#' @method tidy coxreg.multivar
202		#'
203		#' @examples
204		#' multivar_model <- fit_coxreg_multivar(
205		#' variables = list(
206		#' time = "time", event = "status", arm = "armcd",
207		#' covariates = c("covar1", "covar2")
208		#' ),
209		#' data = dta_bladder
210		#' )
211		#' broom::tidy(multivar_model)
212		#'
213		#' @export
214		tidy.coxreg.multivar <- function(x, # nolint
215		...) {
216	16x	checkmate::assert_class(x, "coxreg.multivar")
217	16x	vars <- c(x$vars$arm, x$vars$covariates)
218
219		# Convert the model summaries to data.
220	16x	result <- Map(
221	16x	vars = vars,
222	16x	f = function(vars) {
223	60x	h_coxreg_multivar_extract(
224	60x	var = vars, data = x$data,
225	60x	mod = x$mod, control = x$control
226		)
227		}
228		)
229	16x	result <- do.call(rbind, result)
230
231	16x	result$ci <- Map(lcl = result$lcl, ucl = result$ucl, f = function(lcl, ucl) c(lcl, ucl))
232	16x	result$ci <- lapply(result$ci, empty_vector_if_na)
233	16x	result$hr <- lapply(result$hr, empty_vector_if_na)
234	16x	result$pval <- lapply(result$pval, empty_vector_if_na)
235	16x	result <- result[, names(result) != "n"]
236	16x	attr(result, "conf_level") <- x$control$conf_level
237
238	16x	result
239		}
240
241		#' Fitting functions for Cox proportional hazards regression
242		#'
243		#' @description `r lifecycle::badge("stable")`
244		#'
245		#' Fitting functions for univariate and multivariate Cox regression models.
246		#'
247		#' @param variables (named `list`)\cr the names of the variables found in `data`, passed as a named list and
248		#' corresponding to the `time`, `event`, `arm`, `strata`, and `covariates` terms. If `arm` is missing from
249		#' `variables`, then only Cox model(s) including the `covariates` will be fitted and the corresponding effect
250		#' estimates will be tabulated later.
251		#' @param data (`data.frame`)\cr the dataset containing the variables to fit the models.
252		#' @param at (`list` of `numeric`)\cr when the candidate covariate is a `numeric`, use `at` to specify
253		#' the value of the covariate at which the effect should be estimated.
254		#' @param control (`list`)\cr a list of parameters as returned by the helper function [control_coxreg()].
255		#'
256		#' @seealso [h_cox_regression] for relevant helper functions, [cox_regression].
257		#'
258		#' @examples
259		#' library(survival)
260		#'
261		#' set.seed(1, kind = "Mersenne-Twister")
262		#'
263		#' # Testing dataset [survival::bladder].
264		#' dta_bladder <- with(
265		#' data = bladder[bladder$enum < 5, ],
266		#' data.frame(
267		#' time = stop,
268		#' status = event,
269		#' armcd = as.factor(rx),
270		#' covar1 = as.factor(enum),
271		#' covar2 = factor(
272		#' sample(as.factor(enum)),
273		#' levels = 1:4, labels = c("F", "F", "M", "M")
274		#' )
275		#' )
276		#' )
277		#' labels <- c("armcd" = "ARM", "covar1" = "A Covariate Label", "covar2" = "Sex (F/M)")
278		#' formatters::var_labels(dta_bladder)[names(labels)] <- labels
279		#' dta_bladder$age <- sample(20:60, size = nrow(dta_bladder), replace = TRUE)
280		#'
281		#' plot(
282		#' survfit(Surv(time, status) ~ armcd + covar1, data = dta_bladder),
283		#' lty = 2:4,
284		#' xlab = "Months",
285		#' col = c("blue1", "blue2", "blue3", "blue4", "red1", "red2", "red3", "red4")
286		#' )
287		#'
288		#' @name fit_coxreg
289		NULL
290
291		#' @describeIn fit_coxreg Fit a series of univariate Cox regression models given the inputs.
292		#'
293		#' @return
294		#' * `fit_coxreg_univar()` returns a `coxreg.univar` class object which is a named `list`
295		#' with 5 elements:
296		#' * `mod`: Cox regression models fitted by [survival::coxph()].
297		#' * `data`: The original data frame input.
298		#' * `control`: The original control input.
299		#' * `vars`: The variables used in the model.
300		#' * `at`: Value of the covariate at which the effect should be estimated.
301		#'
302		#' @note When using `fit_coxreg_univar` there should be two study arms.
303		#'
304		#' @examples
305		#' # fit_coxreg_univar
306		#'
307		#' ## Cox regression: arm + 1 covariate.
308		#' mod1 <- fit_coxreg_univar(
309		#' variables = list(
310		#' time = "time", event = "status", arm = "armcd",
311		#' covariates = "covar1"
312		#' ),
313		#' data = dta_bladder,
314		#' control = control_coxreg(conf_level = 0.91)
315		#' )
316		#'
317		#' ## Cox regression: arm + 1 covariate + interaction, 2 candidate covariates.
318		#' mod2 <- fit_coxreg_univar(
319		#' variables = list(
320		#' time = "time", event = "status", arm = "armcd",
321		#' covariates = c("covar1", "covar2")
322		#' ),
323		#' data = dta_bladder,
324		#' control = control_coxreg(conf_level = 0.91, interaction = TRUE)
325		#' )
326		#'
327		#' ## Cox regression: arm + 1 covariate, stratified analysis.
328		#' mod3 <- fit_coxreg_univar(
329		#' variables = list(
330		#' time = "time", event = "status", arm = "armcd", strata = "covar2",
331		#' covariates = c("covar1")
332		#' ),
333		#' data = dta_bladder,
334		#' control = control_coxreg(conf_level = 0.91)
335		#' )
336		#'
337		#' ## Cox regression: no arm, only covariates.
338		#' mod4 <- fit_coxreg_univar(
339		#' variables = list(
340		#' time = "time", event = "status",
341		#' covariates = c("covar1", "covar2")
342		#' ),
343		#' data = dta_bladder
344		#' )
345		#'
346		#' @export
347		fit_coxreg_univar <- function(variables,
348		data,
349		at = list(),
350		control = control_coxreg()) {
351	43x	checkmate::assert_list(variables, names = "named")
352	43x	has_arm <- "arm" %in% names(variables)
353	43x	arm_name <- if (has_arm) "arm" else NULL
354
355	43x	checkmate::assert_character(variables$covariates, null.ok = TRUE)
356
357	43x	assert_df_with_variables(data, variables)
358	43x	assert_list_of_variables(variables[c(arm_name, "event", "time")])
359
360	43x	if (!is.null(variables$strata)) {
361	4x	checkmate::assert_disjunct(control$pval_method, "likelihood")
362		}
363	42x	if (has_arm) {
364	36x	assert_df_with_factors(data, list(val = variables$arm), min.levels = 2, max.levels = 2)
365		}
366	41x	vars <- unlist(variables[c(arm_name, "covariates", "strata")], use.names = FALSE)
367	41x	for (i in vars) {
368	94x	if (is.factor(data[[i]])) {
369	82x	attr(data[[i]], "levels") <- levels(droplevels(data[[i]]))
370		}
371		}
372	41x	forms <- h_coxreg_univar_formulas(variables, interaction = control$interaction)
373	41x	mod <- lapply(
374	41x	forms, function(x) {
375	90x	survival::coxph(formula = stats::as.formula(x), data = data, ties = control$ties)
376		}
377		)
378	41x	structure(
379	41x	list(
380	41x	mod = mod,
381	41x	data = data,
382	41x	control = control,
383	41x	vars = variables,
384	41x	at = at
385		),
386	41x	class = "coxreg.univar"
387		)
388		}
389
390		#' @describeIn fit_coxreg Fit a multivariate Cox regression model.
391		#'
392		#' @return
393		#' * `fit_coxreg_multivar()` returns a `coxreg.multivar` class object which is a named list
394		#' with 4 elements:
395		#' * `mod`: Cox regression model fitted by [survival::coxph()].
396		#' * `data`: The original data frame input.
397		#' * `control`: The original control input.
398		#' * `vars`: The variables used in the model.
399		#'
400		#' @examples
401		#' # fit_coxreg_multivar
402		#'
403		#' ## Cox regression: multivariate Cox regression.
404		#' multivar_model <- fit_coxreg_multivar(
405		#' variables = list(
406		#' time = "time", event = "status", arm = "armcd",
407		#' covariates = c("covar1", "covar2")
408		#' ),
409		#' data = dta_bladder
410		#' )
411		#'
412		#' # Example without treatment arm.
413		#' multivar_covs_model <- fit_coxreg_multivar(
414		#' variables = list(
415		#' time = "time", event = "status",
416		#' covariates = c("covar1", "covar2")
417		#' ),
418		#' data = dta_bladder
419		#' )
420		#'
421		#' @export
422		fit_coxreg_multivar <- function(variables,
423		data,
424		control = control_coxreg()) {
425	83x	checkmate::assert_list(variables, names = "named")
426	83x	has_arm <- "arm" %in% names(variables)
427	83x	arm_name <- if (has_arm) "arm" else NULL
428
429	83x	if (!is.null(variables$covariates)) {
430	21x	checkmate::assert_character(variables$covariates)
431		}
432
433	83x	checkmate::assert_false(control$interaction)
434	83x	assert_df_with_variables(data, variables)
435	83x	assert_list_of_variables(variables[c(arm_name, "event", "time")])
436
437	83x	if (!is.null(variables$strata)) {
438	3x	checkmate::assert_disjunct(control$pval_method, "likelihood")
439		}
440
441	82x	form <- h_coxreg_multivar_formula(variables)
442	82x	mod <- survival::coxph(
443	82x	formula = stats::as.formula(form),
444	82x	data = data,
445	82x	ties = control$ties
446		)
447	82x	structure(
448	82x	list(
449	82x	mod = mod,
450	82x	data = data,
451	82x	control = control,
452	82x	vars = variables
453		),
454	82x	class = "coxreg.multivar"
455		)
456		}
457
458		#' Muffled `car::Anova`
459		#'
460		#' Applied on survival models, [car::Anova()] signal that the `strata` terms is dropped from the model formula when
461		#' present, this function deliberately muffles this message.
462		#'
463		#' @param mod (`coxph`)\cr Cox regression model fitted by [survival::coxph()].
464		#' @param test_statistic (`string`)\cr the method used for estimation of p.values; `wald` (default) or `likelihood`.
465		#'
466		#' @return The output of [car::Anova()], with convergence message muffled.
467		#'
468		#' @keywords internal
469		muffled_car_anova <- function(mod, test_statistic) {
470	219x	tryCatch(
471	219x	withCallingHandlers(
472	219x	expr = {
473	219x	car::Anova(
474	219x	mod,
475	219x	test.statistic = test_statistic,
476	219x	type = "III"
477		)
478		},
479	219x	message = function(m) invokeRestart("muffleMessage"),
480	219x	error = function(e) {
481	1x	stop(paste(
482	1x	"the model seems to have convergence problems, please try to change",
483	1x	"the configuration of covariates or strata variables, e.g.",
484	1x	"- original error:", e
485		))
486		}
487		)
488		)
489		}

1		#' Combine factor levels
2		#'
3		#' @description `r lifecycle::badge("stable")`
4		#'
5		#' Combine specified old factor Levels in a single new level.
6		#'
7		#' @param x (`factor`)\cr factor variable.
8		#' @param levels (`character`)\cr level names to be combined.
9		#' @param new_level (`string`)\cr name of new level.
10		#'
11		#' @return A `factor` with the new levels.
12		#'
13		#' @examples
14		#' x <- factor(letters[1:5], levels = letters[5:1])
15		#' combine_levels(x, levels = c("a", "b"))
16		#'
17		#' combine_levels(x, c("e", "b"))
18		#'
19		#' @export
20		combine_levels <- function(x, levels, new_level = paste(levels, collapse = "/")) {
21	4x	checkmate::assert_factor(x)
22	4x	checkmate::assert_subset(levels, levels(x))
23
24	4x	lvls <- levels(x)
25
26	4x	lvls[lvls %in% levels] <- new_level
27
28	4x	levels(x) <- lvls
29
30	4x	x
31		}
32
33		#' Conversion of a vector to a factor
34		#'
35		#' Converts `x` to a factor and keeps its attributes. Warns appropriately such that the user
36		#' can decide whether they prefer converting to factor manually (e.g. for full control of
37		#' factor levels).
38		#'
39		#' @param x (`vector`)\cr object to convert.
40		#' @param x_name (`string`)\cr name of `x`.
41		#' @param na_level (`string`)\cr the explicit missing level which should be used when converting a character vector.
42		#' @param verbose (`flag`)\cr defaults to `TRUE`. It prints out warnings and messages.
43		#'
44		#' @return A `factor` with same attributes (except class) as `x`. Does not modify `x` if already a `factor`.
45		#'
46		#' @keywords internal
47		as_factor_keep_attributes <- function(x,
48		x_name = deparse(substitute(x)),
49		na_level = "<Missing>",
50		verbose = TRUE) {
51	199x	checkmate::assert_atomic(x)
52	199x	checkmate::assert_string(x_name)
53	199x	checkmate::assert_string(na_level)
54	199x	checkmate::assert_flag(verbose)
55	199x	if (is.factor(x)) {
56	180x	return(x)
57		}
58	19x	x_class <- class(x)[1]
59	19x	if (verbose) {
60	15x	warning(paste(
61	15x	"automatically converting", x_class, "variable", x_name,
62	15x	"to factor, better manually convert to factor to avoid failures"
63		))
64		}
65	19x	if (identical(length(x), 0L)) {
66	1x	warning(paste(
67	1x	x_name, "has length 0, this can lead to tabulation failures, better convert to factor"
68		))
69		}
70	19x	if (is.character(x)) {
71	19x	x_no_na <- explicit_na(sas_na(x), label = na_level)
72	19x	if (any(na_level %in% x_no_na)) {
73	3x	do.call(
74	3x	structure,
75	3x	c(
76	3x	list(.Data = forcats::fct_relevel(x_no_na, na_level, after = Inf)),
77	3x	attributes(x)
78		)
79		)
80		} else {
81	16x	do.call(structure, c(list(.Data = as.factor(x)), attributes(x)))
82		}
83		} else {
84	!	do.call(structure, c(list(.Data = as.factor(x)), attributes(x)))
85		}
86		}
87
88		#' Labels for bins in percent
89		#'
90		#' This creates labels for quantile based bins in percent. This assumes the right-closed
91		#' intervals as produced by [cut_quantile_bins()].
92		#'
93		#' @param probs (`numeric`)\cr the probabilities identifying the quantiles.
94		#' This is a sorted vector of unique `proportion` values, i.e. between 0 and 1, where
95		#' the boundaries 0 and 1 must not be included.
96		#' @param digits (`integer(1)`)\cr number of decimal places to round the percent numbers.
97		#'
98		#' @return A `character` vector with labels in the format `[0%,20%]`, `(20%,50%]`, etc.
99		#'
100		#' @keywords internal
101		bins_percent_labels <- function(probs,
102		digits = 0) {
103	3x	if (isFALSE(0 %in% probs)) probs <- c(0, probs)
104	3x	if (isFALSE(1 %in% probs)) probs <- c(probs, 1)
105	10x	checkmate::assert_numeric(probs, lower = 0, upper = 1, unique = TRUE, sorted = TRUE)
106	10x	percent <- round(probs * 100, digits = digits)
107	10x	left <- paste0(utils::head(percent, -1), "%")
108	10x	right <- paste0(utils::tail(percent, -1), "%")
109	10x	without_left_bracket <- paste0(left, ",", right, "]")
110	10x	with_left_bracket <- paste0("[", utils::head(without_left_bracket, 1))
111	10x	if (length(without_left_bracket) > 1) {
112	7x	with_left_bracket <- c(
113	7x	with_left_bracket,
114	7x	paste0("(", utils::tail(without_left_bracket, -1))
115		)
116		}
117	10x	with_left_bracket
118		}
119
120		#' Cut numeric vector into empirical quantile bins
121		#'
122		#' @description `r lifecycle::badge("stable")`
123		#'
124		#' This cuts a numeric vector into sample quantile bins.
125		#'
126		#' @inheritParams bins_percent_labels
127		#' @param x (`numeric`)\cr the continuous variable values which should be cut into
128		#' quantile bins. This may contain `NA` values, which are then
129		#' not used for the quantile calculations, but included in the return vector.
130		#' @param labels (`character`)\cr the unique labels for the quantile bins. When there are `n`
131		#' probabilities in `probs`, then this must be `n + 1` long.
132		#' @param type (`integer(1)`)\cr type of quantiles to use, see [stats::quantile()] for details.
133		#' @param ordered (`flag`)\cr should the result be an ordered factor.
134		#'
135		#' @return A `factor` variable with appropriately-labeled bins as levels.
136		#'
137		#' @note Intervals are closed on the right side. That is, the first bin is the interval
138		#' `[-Inf, q1]` where `q1` is the first quantile, the second bin is then `(q1, q2]`, etc.,
139		#' and the last bin is `(qn, +Inf]` where `qn` is the last quantile.
140		#'
141		#' @examples
142		#' # Default is to cut into quartile bins.
143		#' cut_quantile_bins(cars$speed)
144		#'
145		#' # Use custom quantiles.
146		#' cut_quantile_bins(cars$speed, probs = c(0.1, 0.2, 0.6, 0.88))
147		#'
148		#' # Use custom labels.
149		#' cut_quantile_bins(cars$speed, labels = paste0("Q", 1:4))
150		#'
151		#' # NAs are preserved in result factor.
152		#' ozone_binned <- cut_quantile_bins(airquality$Ozone)
153		#' which(is.na(ozone_binned))
154		#' # So you might want to make these explicit.
155		#' explicit_na(ozone_binned)
156		#'
157		#' @export
158		cut_quantile_bins <- function(x,
159		probs = c(0.25, 0.5, 0.75),
160		labels = NULL,
161		type = 7,
162		ordered = TRUE) {
163	8x	checkmate::assert_flag(ordered)
164	8x	checkmate::assert_numeric(x)
165	7x	if (isFALSE(0 %in% probs)) probs <- c(0, probs)
166	7x	if (isFALSE(1 %in% probs)) probs <- c(probs, 1)
167	8x	checkmate::assert_numeric(probs, lower = 0, upper = 1, unique = TRUE, sorted = TRUE)
168	7x	if (is.null(labels)) labels <- bins_percent_labels(probs)
169	8x	checkmate::assert_character(labels, len = length(probs) - 1, any.missing = FALSE, unique = TRUE)
170
171	8x	if (all(is.na(x))) {
172		# Early return if there are only NAs in input.
173	1x	return(factor(x, ordered = ordered, levels = labels))
174		}
175
176	7x	quantiles <- stats::quantile(
177	7x	x,
178	7x	probs = probs,
179	7x	type = type,
180	7x	na.rm = TRUE
181		)
182
183	7x	checkmate::assert_numeric(quantiles, unique = TRUE)
184
185	6x	cut(
186	6x	x,
187	6x	breaks = quantiles,
188	6x	labels = labels,
189	6x	ordered_result = ordered,
190	6x	include.lowest = TRUE,
191	6x	right = TRUE
192		)
193		}
194
195		#' Discard specified levels of a factor
196		#'
197		#' @description `r lifecycle::badge("stable")`
198		#'
199		#' This discards the observations as well as the levels specified from a factor.
200		#'
201		#' @param x (`factor`)\cr the original factor.
202		#' @param discard (`character`)\cr levels to discard.
203		#'
204		#' @return A modified `factor` with observations as well as levels from `discard` dropped.
205		#'
206		#' @examples
207		#' fct_discard(factor(c("a", "b", "c")), "c")
208		#'
209		#' @export
210		fct_discard <- function(x, discard) {
211	319x	checkmate::assert_factor(x)
212	319x	checkmate::assert_character(discard, any.missing = FALSE)
213	319x	new_obs <- x[!(x %in% discard)]
214	319x	new_levels <- setdiff(levels(x), discard)
215	319x	factor(new_obs, levels = new_levels)
216		}
217
218		#' Insertion of explicit missing values in a factor
219		#'
220		#' @description `r lifecycle::badge("stable")`
221		#'
222		#' This inserts explicit missing values in a factor based on a condition. Additionally,
223		#' existing `NA` values will be explicitly converted to given `na_level`.
224		#'
225		#' @param x (`factor`)\cr the original factor.
226		#' @param condition (`logical`)\cr positions at which to insert missing values.
227		#' @param na_level (`string`)\cr which level to use for missing values.
228		#'
229		#' @return A modified `factor` with inserted and existing `NA` converted to `na_level`.
230		#'
231		#' @seealso [forcats::fct_na_value_to_level()] which is used internally.
232		#'
233		#' @examples
234		#' fct_explicit_na_if(factor(c("a", "b", NA)), c(TRUE, FALSE, FALSE))
235		#'
236		#' @export
237		fct_explicit_na_if <- function(x, condition, na_level = "<Missing>") {
238	1x	checkmate::assert_factor(x, len = length(condition))
239	1x	checkmate::assert_logical(condition)
240	1x	x[condition] <- NA
241	1x	x <- forcats::fct_na_value_to_level(x, level = na_level)
242	1x	forcats::fct_drop(x, only = na_level)
243		}
244
245		#' Collapse factor levels and keep only those new group levels
246		#'
247		#' @description `r lifecycle::badge("stable")`
248		#'
249		#' This collapses levels and only keeps those new group levels, in the order provided.
250		#' The returned factor has levels in the order given, with the possible missing level last (this will
251		#' only be included if there are missing values).
252		#'
253		#' @param .f (`factor` or `character`)\cr original vector.
254		#' @param ... (named `character`)\cr levels in each vector provided will be collapsed into
255		#' the new level given by the respective name.
256		#' @param .na_level (`string`)\cr which level to use for other levels, which should be missing in the
257		#' new factor. Note that this level must not be contained in the new levels specified in `...`.
258		#'
259		#' @return A modified `factor` with collapsed levels. Values and levels which are not included
260		#' in the given `character` vector input will be set to the missing level `.na_level`.
261		#'
262		#' @note Any existing `NA`s in the input vector will not be replaced by the missing level. If needed,
263		#' [explicit_na()] can be called separately on the result.
264		#'
265		#' @seealso [forcats::fct_collapse()], [forcats::fct_relevel()] which are used internally.
266		#'
267		#' @examples
268		#' fct_collapse_only(factor(c("a", "b", "c", "d")), TRT = "b", CTRL = c("c", "d"))
269		#'
270		#' @export
271		fct_collapse_only <- function(.f, ..., .na_level = "<Missing>") {
272	4x	new_lvls <- names(list(...))
273	4x	if (checkmate::test_subset(.na_level, new_lvls)) {
274	1x	stop(paste0(".na_level currently set to '", .na_level, "' must not be contained in the new levels"))
275		}
276	3x	x <- forcats::fct_collapse(.f, ..., other_level = .na_level)
277	3x	do.call(forcats::fct_relevel, args = c(list(.f = x), as.list(new_lvls)))
278		}
279
280		#' Ungroup non-numeric statistics
281		#'
282		#' Ungroups grouped non-numeric statistics within input vectors `.formats`, `.labels`, and `.indent_mods`.
283		#'
284		#' @inheritParams argument_convention
285		#' @param x (named `list` of `numeric`)\cr list of numeric statistics containing the statistics to ungroup.
286		#'
287		#' @return A `list` with modified elements `x`, `.formats`, `.labels`, and `.indent_mods`.
288		#'
289		#' @seealso [a_summary()] which uses this function internally.
290		#'
291		#' @keywords internal
292		ungroup_stats <- function(x,
293		.formats,
294		.labels,
295		.indent_mods) {
296	316x	checkmate::assert_list(x)
297	316x	empty_pval <- "pval" %in% names(x) && length(x[["pval"]]) == 0
298	316x	empty_pval_counts <- "pval_counts" %in% names(x) && length(x[["pval_counts"]]) == 0
299	316x	x <- unlist(x, recursive = FALSE)
300
301		# If p-value is empty it is removed by unlist and needs to be re-added
302	!	if (empty_pval) x[["pval"]] <- character()
303	3x	if (empty_pval_counts) x[["pval_counts"]] <- character()
304	316x	.stats <- names(x)
305
306		# Ungroup stats
307	316x	.formats <- lapply(.stats, function(x) {
308	2644x	.formats[[if (!grepl("\\.", x)) x else regmatches(x, regexpr("\\.", x), invert = TRUE)[[1]][1]]]
309		})
310	316x	.indent_mods <- sapply(.stats, function(x) {
311	2644x	.indent_mods[[if (!grepl("\\.", x)) x else regmatches(x, regexpr("\\.", x), invert = TRUE)[[1]][1]]]
312		})
313	316x	.labels <- sapply(.stats, function(x) {
314	2575x	if (!grepl("\\.", x)) .labels[[x]] else regmatches(x, regexpr("\\.", x), invert = TRUE)[[1]][2]
315		})
316
317	316x	list(
318	316x	x = x,
319	316x	.formats = .formats,
320	316x	.labels = .labels,
321	316x	.indent_mods = .indent_mods
322		)
323		}

1		#' Tabulate binary response by subgroup
2		#'
3		#' @description `r lifecycle::badge("stable")`
4		#'
5		#' The [tabulate_rsp_subgroups()] function creates a layout element to tabulate binary response by subgroup, returning
6		#' statistics including response rate and odds ratio for each population subgroup. The table is created from `df`, a
7		#' list of data frames returned by [extract_rsp_subgroups()], with the statistics to include specified via the `vars`
8		#' parameter.
9		#'
10		#' A forest plot can be created from the resulting table using the [g_forest()] function.
11		#'
12		#' @inheritParams extract_rsp_subgroups
13		#' @inheritParams argument_convention
14		#'
15		#' @details These functions create a layout starting from a data frame which contains
16		#' the required statistics. Tables typically used as part of forest plot.
17		#'
18		#' @seealso [extract_rsp_subgroups()]
19		#'
20		#' @examples
21		#' library(dplyr)
22		#' library(forcats)
23		#'
24		#' adrs <- tern_ex_adrs
25		#' adrs_labels <- formatters::var_labels(adrs)
26		#'
27		#' adrs_f <- adrs %>%
28		#' filter(PARAMCD == "BESRSPI") %>%
29		#' filter(ARM %in% c("A: Drug X", "B: Placebo")) %>%
30		#' droplevels() %>%
31		#' mutate(
32		#' # Reorder levels of factor to make the placebo group the reference arm.
33		#' ARM = fct_relevel(ARM, "B: Placebo"),
34		#' rsp = AVALC == "CR"
35		#' )
36		#' formatters::var_labels(adrs_f) <- c(adrs_labels, "Response")
37		#'
38		#' # Unstratified analysis.
39		#' df <- extract_rsp_subgroups(
40		#' variables = list(rsp = "rsp", arm = "ARM", subgroups = c("SEX", "BMRKR2")),
41		#' data = adrs_f
42		#' )
43		#' df
44		#'
45		#' # Stratified analysis.
46		#' df_strat <- extract_rsp_subgroups(
47		#' variables = list(rsp = "rsp", arm = "ARM", subgroups = c("SEX", "BMRKR2"), strata = "STRATA1"),
48		#' data = adrs_f
49		#' )
50		#' df_strat
51		#'
52		#' # Grouping of the BMRKR2 levels.
53		#' df_grouped <- extract_rsp_subgroups(
54		#' variables = list(rsp = "rsp", arm = "ARM", subgroups = c("SEX", "BMRKR2")),
55		#' data = adrs_f,
56		#' groups_lists = list(
57		#' BMRKR2 = list(
58		#' "low" = "LOW",
59		#' "low/medium" = c("LOW", "MEDIUM"),
60		#' "low/medium/high" = c("LOW", "MEDIUM", "HIGH")
61		#' )
62		#' )
63		#' )
64		#' df_grouped
65		#'
66		#' @name response_subgroups
67		#' @order 1
68		NULL
69
70		#' Prepare response data for population subgroups in data frames
71		#'
72		#' @description `r lifecycle::badge("stable")`
73		#'
74		#' Prepares response rates and odds ratios for population subgroups in data frames. Simple wrapper
75		#' for [h_odds_ratio_subgroups_df()] and [h_proportion_subgroups_df()]. Result is a list of two
76		#' `data.frames`: `prop` and `or`. `variables` corresponds to the names of variables found in `data`,
77		#' passed as a named `list` and requires elements `rsp`, `arm` and optionally `subgroups` and `strata`.
78		#' `groups_lists` optionally specifies groupings for `subgroups` variables.
79		#'
80		#' @inheritParams argument_convention
81		#' @inheritParams response_subgroups
82		#' @param label_all (`string`)\cr label for the total population analysis.
83		#'
84		#' @return A named list of two elements:
85		#' * `prop`: A `data.frame` containing columns `arm`, `n`, `n_rsp`, `prop`, `subgroup`, `var`,
86		#' `var_label`, and `row_type`.
87		#' * `or`: A `data.frame` containing columns `arm`, `n_tot`, `or`, `lcl`, `ucl`, `conf_level`,
88		#' `subgroup`, `var`, `var_label`, and `row_type`.
89		#'
90		#' @seealso [response_subgroups]
91		#'
92		#' @export
93		extract_rsp_subgroups <- function(variables,
94		data,
95		groups_lists = list(),
96		conf_level = 0.95,
97		method = NULL,
98		label_all = "All Patients") {
99	14x	if ("strat" %in% names(variables)) {
100	!	warning(
101	!	"Warning: the `strat` element name of the `variables` list argument to `extract_rsp_subgroups() ",
102	!	"was deprecated in tern 0.9.4.\n ",
103	!	"Please use the name `strata` instead of `strat` in the `variables` argument."
104		)
105	!	variables[["strata"]] <- variables[["strat"]]
106		}
107
108	14x	df_prop <- h_proportion_subgroups_df(
109	14x	variables,
110	14x	data,
111	14x	groups_lists = groups_lists,
112	14x	label_all = label_all
113		)
114	14x	df_or <- h_odds_ratio_subgroups_df(
115	14x	variables,
116	14x	data,
117	14x	groups_lists = groups_lists,
118	14x	conf_level = conf_level,
119	14x	method = method,
120	14x	label_all = label_all
121		)
122
123	14x	list(prop = df_prop, or = df_or)
124		}
125
126		#' @describeIn response_subgroups Formatted analysis function which is used as `afun` in `tabulate_rsp_subgroups()`.
127		#'
128		#' @return
129		#' * `a_response_subgroups()` returns the corresponding list with formatted [rtables::CellValue()].
130		#'
131		#' @keywords internal
132		a_response_subgroups <- function(.formats = list(
133		n = "xx", # nolint start
134		n_rsp = "xx",
135		prop = "xx.x%",
136		n_tot = "xx",
137		or = list(format_extreme_values(2L)),
138		ci = list(format_extreme_values_ci(2L)),
139		pval = "x.xxxx \| (<0.0001)",
140		riskdiff = "xx.x (xx.x - xx.x)" # nolint end
141		),
142		na_str = default_na_str()) {
143	22x	checkmate::assert_list(.formats)
144	22x	checkmate::assert_subset(
145	22x	names(.formats),
146	22x	c("n", "n_rsp", "prop", "n_tot", "or", "ci", "pval", "riskdiff")
147		)
148
149	22x	afun_lst <- Map(
150	22x	function(stat, fmt, na_str) {
151	157x	function(df, labelstr = "", ...) {
152	349x	in_rows(
153	349x	.list = as.list(df[[stat]]),
154	349x	.labels = as.character(df$subgroup),
155	349x	.formats = fmt,
156	349x	.format_na_strs = na_str
157		)
158		}
159		},
160	22x	stat = names(.formats),
161	22x	fmt = .formats,
162	22x	na_str = na_str
163		)
164
165	22x	afun_lst
166		}
167
168		#' @describeIn response_subgroups Table-creating function which creates a table
169		#' summarizing binary response by subgroup. This function is a wrapper for [rtables::analyze_colvars()]
170		#' and [rtables::summarize_row_groups()].
171		#'
172		#' @param df (`list`)\cr a list of data frames containing all analysis variables. List should be
173		#' created using [extract_rsp_subgroups()].
174		#' @param vars (`character`)\cr the names of statistics to be reported among:
175		#' * `n`: Total number of observations per group.
176		#' * `n_rsp`: Number of responders per group.
177		#' * `prop`: Proportion of responders.
178		#' * `n_tot`: Total number of observations.
179		#' * `or`: Odds ratio.
180		#' * `ci` : Confidence interval of odds ratio.
181		#' * `pval`: p-value of the effect.
182		#' Note, the statistics `n_tot`, `or`, and `ci` are required.
183		#' @param riskdiff (`list`)\cr if a risk (proportion) difference column should be added, a list of settings to apply
184		#' within the column. See [control_riskdiff()] for details. If `NULL`, no risk difference column will be added. If
185		#' `riskdiff$arm_x` and `riskdiff$arm_y` are `NULL`, the first level of `df$prop$arm` will be used as `arm_x` and
186		#' the second level as `arm_y`.
187		#'
188		#' @return An `rtables` table summarizing binary response by subgroup.
189		#'
190		#' @examples
191		#' # Table with default columns
192		#' basic_table() %>%
193		#' tabulate_rsp_subgroups(df)
194		#'
195		#' # Table with selected columns
196		#' basic_table() %>%
197		#' tabulate_rsp_subgroups(
198		#' df = df,
199		#' vars = c("n_tot", "n", "n_rsp", "prop", "or", "ci")
200		#' )
201		#'
202		#' # Table with risk difference column added
203		#' basic_table() %>%
204		#' tabulate_rsp_subgroups(
205		#' df,
206		#' riskdiff = control_riskdiff(
207		#' arm_x = levels(df$prop$arm)[1],
208		#' arm_y = levels(df$prop$arm)[2]
209		#' )
210		#' )
211		#'
212		#' @export
213		#' @order 2
214		tabulate_rsp_subgroups <- function(lyt,
215		df,
216		vars = c("n_tot", "n", "prop", "or", "ci"),
217		groups_lists = list(),
218		label_all = "All Patients",
219		riskdiff = NULL,
220		na_str = default_na_str(),
221		.formats = c(
222		n = "xx", n_rsp = "xx", prop = "xx.x%", n_tot = "xx",
223		or = list(format_extreme_values(2L)), ci = list(format_extreme_values_ci(2L)),
224		pval = "x.xxxx \| (<0.0001)"
225		)) {
226	13x	checkmate::assert_list(riskdiff, null.ok = TRUE)
227	13x	checkmate::assert_true(all(c("n_tot", "or", "ci") %in% vars))
228	13x	if ("pval" %in% vars && !"pval" %in% names(df$or)) {
229	1x	warning(
230	1x	'The "pval" statistic has been selected but is not present in "df" so it will not be included in the output ',
231	1x	'table. To include the "pval" statistic, please specify a p-value test when generating "df" via ',
232	1x	'the "method" argument to `extract_rsp_subgroups()`. If method = "cmh", strata must also be specified via the ',
233	1x	'"variables" argument to `extract_rsp_subgroups()`.'
234		)
235		}
236
237		# Create "ci" column from "lcl" and "ucl"
238	13x	df$or$ci <- combine_vectors(df$or$lcl, df$or$ucl)
239
240		# Fill in missing formats with defaults
241	13x	default_fmts <- eval(formals(tabulate_rsp_subgroups)$.formats)
242	13x	.formats <- c(.formats, default_fmts[vars[!vars %in% names(.formats)]])
243
244		# Extract additional parameters from df
245	13x	conf_level <- df$or$conf_level[1]
246	13x	method <- if ("pval_label" %in% names(df$or)) df$or$pval_label[1] else NULL
247	13x	colvars <- d_rsp_subgroups_colvars(vars, conf_level = conf_level, method = method)
248	13x	prop_vars <- intersect(colvars$vars, c("n", "prop", "n_rsp"))
249	13x	or_vars <- intersect(names(colvars$labels), c("n_tot", "or", "ci", "pval"))
250	13x	colvars_prop <- list(vars = prop_vars, labels = colvars$labels[prop_vars])
251	13x	colvars_or <- list(vars = or_vars, labels = colvars$labels[or_vars])
252
253	13x	extra_args <- list(groups_lists = groups_lists, conf_level = conf_level, method = method, label_all = label_all)
254
255		# Get analysis function for each statistic
256	13x	afun_lst <- a_response_subgroups(.formats = c(.formats, riskdiff = riskdiff$format), na_str = na_str)
257
258		# Add risk difference column
259	13x	if (!is.null(riskdiff)) {
260	!	if (is.null(riskdiff$arm_x)) riskdiff$arm_x <- levels(df$prop$arm)[1]
261	!	if (is.null(riskdiff$arm_y)) riskdiff$arm_y <- levels(df$prop$arm)[2]
262	1x	colvars_or$vars <- c(colvars_or$vars, "riskdiff")
263	1x	colvars_or$labels <- c(colvars_or$labels, riskdiff = riskdiff$col_label)
264	1x	arm_cols <- paste(rep(c("n_rsp", "n_rsp", "n", "n")), c(riskdiff$arm_x, riskdiff$arm_y), sep = "_")
265
266	1x	df_prop_diff <- df$prop %>%
267	1x	dplyr::select(-"prop") %>%
268	1x	tidyr::pivot_wider(
269	1x	id_cols = c("subgroup", "var", "var_label", "row_type"),
270	1x	names_from = "arm",
271	1x	values_from = c("n", "n_rsp")
272		) %>%
273	1x	dplyr::rowwise() %>%
274	1x	dplyr::mutate(
275	1x	riskdiff = stat_propdiff_ci(
276	1x	x = as.list(.data[[arm_cols[1]]]),
277	1x	y = as.list(.data[[arm_cols[2]]]),
278	1x	N_x = .data[[arm_cols[3]]],
279	1x	N_y = .data[[arm_cols[4]]]
280		)
281		) %>%
282	1x	dplyr::select(-dplyr::all_of(arm_cols))
283
284	1x	df$or <- df$or %>%
285	1x	dplyr::left_join(
286	1x	df_prop_diff,
287	1x	by = c("subgroup", "var", "var_label", "row_type")
288		)
289		}
290
291		# Add columns from table_prop (optional)
292	13x	if (length(colvars_prop$vars) > 0) {
293	12x	lyt_prop <- split_cols_by(lyt = lyt, var = "arm")
294	12x	lyt_prop <- split_cols_by_multivar(
295	12x	lyt = lyt_prop,
296	12x	vars = colvars_prop$vars,
297	12x	varlabels = colvars_prop$labels
298		)
299
300		# Add "All Patients" row
301	12x	lyt_prop <- split_rows_by(
302	12x	lyt = lyt_prop,
303	12x	var = "row_type",
304	12x	split_fun = keep_split_levels("content"),
305	12x	nested = FALSE,
306	12x	child_labels = "hidden"
307		)
308	12x	lyt_prop <- analyze_colvars(
309	12x	lyt = lyt_prop,
310	12x	afun = afun_lst[names(colvars_prop$labels)],
311	12x	na_str = na_str,
312	12x	extra_args = extra_args
313		)
314
315		# Add analysis rows
316	12x	if ("analysis" %in% df$prop$row_type) {
317	11x	lyt_prop <- split_rows_by(
318	11x	lyt = lyt_prop,
319	11x	var = "row_type",
320	11x	split_fun = keep_split_levels("analysis"),
321	11x	nested = FALSE,
322	11x	child_labels = "hidden"
323		)
324	11x	lyt_prop <- split_rows_by(lyt = lyt_prop, var = "var_label", nested = TRUE)
325	11x	lyt_prop <- analyze_colvars(
326	11x	lyt = lyt_prop,
327	11x	afun = afun_lst[names(colvars_prop$labels)],
328	11x	na_str = na_str,
329	11x	inclNAs = TRUE,
330	11x	extra_args = extra_args
331		)
332		}
333
334	12x	table_prop <- build_table(lyt_prop, df = df$prop)
335		} else {
336	1x	table_prop <- NULL
337		}
338
339		# Add columns from table_or ("n_tot", "or", and "ci" required)
340	13x	lyt_or <- split_cols_by(lyt = lyt, var = "arm")
341	13x	lyt_or <- split_cols_by_multivar(
342	13x	lyt = lyt_or,
343	13x	vars = colvars_or$vars,
344	13x	varlabels = colvars_or$labels
345		)
346
347		# Add "All Patients" row
348	13x	lyt_or <- split_rows_by(
349	13x	lyt = lyt_or,
350	13x	var = "row_type",
351	13x	split_fun = keep_split_levels("content"),
352	13x	nested = FALSE,
353	13x	child_labels = "hidden"
354		)
355	13x	lyt_or <- analyze_colvars(
356	13x	lyt = lyt_or,
357	13x	afun = afun_lst[names(colvars_or$labels)],
358	13x	na_str = na_str,
359	13x	extra_args = extra_args
360		) %>%
361	13x	append_topleft("Baseline Risk Factors")
362
363		# Add analysis rows
364	13x	if ("analysis" %in% df$or$row_type) {
365	12x	lyt_or <- split_rows_by(
366	12x	lyt = lyt_or,
367	12x	var = "row_type",
368	12x	split_fun = keep_split_levels("analysis"),
369	12x	nested = FALSE,
370	12x	child_labels = "hidden"
371		)
372	12x	lyt_or <- split_rows_by(lyt = lyt_or, var = "var_label", nested = TRUE)
373	12x	lyt_or <- analyze_colvars(
374	12x	lyt = lyt_or,
375	12x	afun = afun_lst[names(colvars_or$labels)],
376	12x	na_str = na_str,
377	12x	inclNAs = TRUE,
378	12x	extra_args = extra_args
379		)
380		}
381
382	13x	table_or <- build_table(lyt_or, df = df$or)
383
384		# Join tables, add forest plot attributes
385	13x	n_tot_id <- match("n_tot", colvars_or$vars)
386	13x	if (is.null(table_prop)) {
387	1x	result <- table_or
388	1x	or_id <- match("or", colvars_or$vars)
389	1x	ci_id <- match("ci", colvars_or$vars)
390		} else {
391	12x	result <- cbind_rtables(table_or[, n_tot_id], table_prop, table_or[, -n_tot_id])
392	12x	or_id <- 1L + ncol(table_prop) + match("or", colvars_or$vars[-n_tot_id])
393	12x	ci_id <- 1L + ncol(table_prop) + match("ci", colvars_or$vars[-n_tot_id])
394	12x	n_tot_id <- 1L
395		}
396	13x	structure(
397	13x	result,
398	13x	forest_header = paste0(levels(df$prop$arm), "\nBetter"),
399	13x	col_x = or_id,
400	13x	col_ci = ci_id,
401	13x	col_symbol_size = n_tot_id
402		)
403		}
404
405		#' Labels for column variables in binary response by subgroup table
406		#'
407		#' @description `r lifecycle::badge("stable")`
408		#'
409		#' Internal function to check variables included in [tabulate_rsp_subgroups()] and create column labels.
410		#'
411		#' @inheritParams argument_convention
412		#' @inheritParams tabulate_rsp_subgroups
413		#'
414		#' @return A `list` of variables to tabulate and their labels.
415		#'
416		#' @export
417		d_rsp_subgroups_colvars <- function(vars,
418		conf_level = NULL,
419		method = NULL) {
420	22x	checkmate::assert_character(vars)
421	22x	checkmate::assert_subset(c("n_tot", "or", "ci"), vars)
422	22x	checkmate::assert_subset(
423	22x	vars,
424	22x	c("n", "n_rsp", "prop", "n_tot", "or", "ci", "pval")
425		)
426
427	22x	varlabels <- c(
428	22x	n = "n",
429	22x	n_rsp = "Responders",
430	22x	prop = "Response (%)",
431	22x	n_tot = "Total n",
432	22x	or = "Odds Ratio"
433		)
434	22x	colvars <- vars
435
436	22x	if ("ci" %in% colvars) {
437	22x	checkmate::assert_false(is.null(conf_level))
438
439	22x	varlabels <- c(
440	22x	varlabels,
441	22x	ci = paste0(100 * conf_level, "% CI")
442		)
443
444		# The `lcl`` variable is just a placeholder available in the analysis data,
445		# it is not acutally used in the tabulation.
446		# Variables used in the tabulation are lcl and ucl, see `a_response_subgroups` for details.
447	22x	colvars[colvars == "ci"] <- "lcl"
448		}
449
450	22x	if ("pval" %in% colvars) {
451	16x	varlabels <- c(
452	16x	varlabels,
453	16x	pval = method
454		)
455		}
456
457	22x	list(
458	22x	vars = colvars,
459	22x	labels = varlabels[vars]
460		)
461		}

1		#' Helper functions for tabulating survival duration by subgroup
2		#'
3		#' @description `r lifecycle::badge("stable")`
4		#'
5		#' Helper functions that tabulate in a data frame statistics such as median survival
6		#' time and hazard ratio for population subgroups.
7		#'
8		#' @inheritParams argument_convention
9		#' @inheritParams survival_coxph_pairwise
10		#' @inheritParams survival_duration_subgroups
11		#' @param arm (`factor`)\cr the treatment group variable.
12		#'
13		#' @details Main functionality is to prepare data for use in a layout-creating function.
14		#'
15		#' @examples
16		#' library(dplyr)
17		#' library(forcats)
18		#'
19		#' adtte <- tern_ex_adtte
20		#'
21		#' # Save variable labels before data processing steps.
22		#' adtte_labels <- formatters::var_labels(adtte)
23		#'
24		#' adtte_f <- adtte %>%
25		#' filter(
26		#' PARAMCD == "OS",
27		#' ARM %in% c("B: Placebo", "A: Drug X"),
28		#' SEX %in% c("M", "F")
29		#' ) %>%
30		#' mutate(
31		#' # Reorder levels of ARM to display reference arm before treatment arm.
32		#' ARM = droplevels(fct_relevel(ARM, "B: Placebo")),
33		#' SEX = droplevels(SEX),
34		#' is_event = CNSR == 0
35		#' )
36		#' labels <- c("ARM" = adtte_labels[["ARM"]], "SEX" = adtte_labels[["SEX"]], "is_event" = "Event Flag")
37		#' formatters::var_labels(adtte_f)[names(labels)] <- labels
38		#'
39		#' @name h_survival_duration_subgroups
40		NULL
41
42		#' @describeIn h_survival_duration_subgroups Helper to prepare a data frame of median survival times by arm.
43		#'
44		#' @return
45		#' * `h_survtime_df()` returns a `data.frame` with columns `arm`, `n`, `n_events`, and `median`.
46		#'
47		#' @examples
48		#' # Extract median survival time for one group.
49		#' h_survtime_df(
50		#' tte = adtte_f$AVAL,
51		#' is_event = adtte_f$is_event,
52		#' arm = adtte_f$ARM
53		#' )
54		#'
55		#' @export
56		h_survtime_df <- function(tte, is_event, arm) {
57	79x	checkmate::assert_numeric(tte)
58	78x	checkmate::assert_logical(is_event, len = length(tte))
59	78x	assert_valid_factor(arm, len = length(tte))
60
61	78x	df_tte <- data.frame(
62	78x	tte = tte,
63	78x	is_event = is_event,
64	78x	stringsAsFactors = FALSE
65		)
66
67		# Delete NAs
68	78x	non_missing_rows <- stats::complete.cases(df_tte)
69	78x	df_tte <- df_tte[non_missing_rows, ]
70	78x	arm <- arm[non_missing_rows]
71
72	78x	lst_tte <- split(df_tte, arm)
73	78x	lst_results <- Map(function(x, arm) {
74	156x	if (nrow(x) > 0) {
75	152x	s_surv <- s_surv_time(x, .var = "tte", is_event = "is_event")
76	152x	median_est <- unname(as.numeric(s_surv$median))
77	152x	n_events <- sum(x$is_event)
78		} else {
79	4x	median_est <- NA
80	4x	n_events <- NA
81		}
82
83	156x	data.frame(
84	156x	arm = arm,
85	156x	n = nrow(x),
86	156x	n_events = n_events,
87	156x	median = median_est,
88	156x	stringsAsFactors = FALSE
89		)
90	78x	}, lst_tte, names(lst_tte))
91
92	78x	df <- do.call(rbind, args = c(lst_results, make.row.names = FALSE))
93	78x	df$arm <- factor(df$arm, levels = levels(arm))
94	78x	df
95		}
96
97		#' @describeIn h_survival_duration_subgroups Summarizes median survival times by arm and across subgroups
98		#' in a data frame. `variables` corresponds to the names of variables found in `data`, passed as a named list and
99		#' requires elements `tte`, `is_event`, `arm` and optionally `subgroups`. `groups_lists` optionally specifies
100		#' groupings for `subgroups` variables.
101		#'
102		#' @return
103		#' * `h_survtime_subgroups_df()` returns a `data.frame` with columns `arm`, `n`, `n_events`, `median`, `subgroup`,
104		#' `var`, `var_label`, and `row_type`.
105		#'
106		#' @examples
107		#' # Extract median survival time for multiple groups.
108		#' h_survtime_subgroups_df(
109		#' variables = list(
110		#' tte = "AVAL",
111		#' is_event = "is_event",
112		#' arm = "ARM",
113		#' subgroups = c("SEX", "BMRKR2")
114		#' ),
115		#' data = adtte_f
116		#' )
117		#'
118		#' # Define groupings for BMRKR2 levels.
119		#' h_survtime_subgroups_df(
120		#' variables = list(
121		#' tte = "AVAL",
122		#' is_event = "is_event",
123		#' arm = "ARM",
124		#' subgroups = c("SEX", "BMRKR2")
125		#' ),
126		#' data = adtte_f,
127		#' groups_lists = list(
128		#' BMRKR2 = list(
129		#' "low" = "LOW",
130		#' "low/medium" = c("LOW", "MEDIUM"),
131		#' "low/medium/high" = c("LOW", "MEDIUM", "HIGH")
132		#' )
133		#' )
134		#' )
135		#'
136		#' @export
137		h_survtime_subgroups_df <- function(variables,
138		data,
139		groups_lists = list(),
140		label_all = "All Patients") {
141	15x	checkmate::assert_character(variables$tte)
142	15x	checkmate::assert_character(variables$is_event)
143	15x	checkmate::assert_character(variables$arm)
144	15x	checkmate::assert_character(variables$subgroups, null.ok = TRUE)
145
146	15x	assert_df_with_variables(data, variables)
147
148	15x	checkmate::assert_string(label_all)
149
150		# Add All Patients.
151	15x	result_all <- h_survtime_df(data[[variables$tte]], data[[variables$is_event]], data[[variables$arm]])
152	15x	result_all$subgroup <- label_all
153	15x	result_all$var <- "ALL"
154	15x	result_all$var_label <- label_all
155	15x	result_all$row_type <- "content"
156
157		# Add Subgroups.
158	15x	if (is.null(variables$subgroups)) {
159	3x	result_all
160		} else {
161	12x	l_data <- h_split_by_subgroups(data, variables$subgroups, groups_lists = groups_lists)
162	12x	l_result <- lapply(l_data, function(grp) {
163	60x	result <- h_survtime_df(grp$df[[variables$tte]], grp$df[[variables$is_event]], grp$df[[variables$arm]])
164	60x	result_labels <- grp$df_labels[rep(1, times = nrow(result)), ]
165	60x	cbind(result, result_labels)
166		})
167	12x	result_subgroups <- do.call(rbind, args = c(l_result, make.row.names = FALSE))
168	12x	result_subgroups$row_type <- "analysis"
169	12x	rbind(
170	12x	result_all,
171	12x	result_subgroups
172		)
173		}
174		}
175
176		#' @describeIn h_survival_duration_subgroups Helper to prepare a data frame with estimates of
177		#' treatment hazard ratio.
178		#'
179		#' @param strata_data (`factor`, `data.frame`, or `NULL`)\cr required if stratified analysis is performed.
180		#'
181		#' @return
182		#' * `h_coxph_df()` returns a `data.frame` with columns `arm`, `n_tot`, `n_tot_events`, `hr`, `lcl`, `ucl`,
183		#' `conf_level`, `pval` and `pval_label`.
184		#'
185		#' @examples
186		#' # Extract hazard ratio for one group.
187		#' h_coxph_df(adtte_f$AVAL, adtte_f$is_event, adtte_f$ARM)
188		#'
189		#' # Extract hazard ratio for one group with stratification factor.
190		#' h_coxph_df(adtte_f$AVAL, adtte_f$is_event, adtte_f$ARM, strata_data = adtte_f$STRATA1)
191		#'
192		#' @export
193		h_coxph_df <- function(tte, is_event, arm, strata_data = NULL, control = control_coxph()) {
194	85x	checkmate::assert_numeric(tte)
195	85x	checkmate::assert_logical(is_event, len = length(tte))
196	85x	assert_valid_factor(arm, n.levels = 2, len = length(tte))
197
198	85x	df_tte <- data.frame(tte = tte, is_event = is_event)
199	85x	strata_vars <- NULL
200
201	85x	if (!is.null(strata_data)) {
202	5x	if (is.data.frame(strata_data)) {
203	4x	strata_vars <- names(strata_data)
204	4x	checkmate::assert_data_frame(strata_data, nrows = nrow(df_tte))
205	4x	assert_df_with_factors(strata_data, as.list(stats::setNames(strata_vars, strata_vars)))
206		} else {
207	1x	assert_valid_factor(strata_data, len = nrow(df_tte))
208	1x	strata_vars <- "strata_data"
209		}
210	5x	df_tte[strata_vars] <- strata_data
211		}
212
213	85x	l_df <- split(df_tte, arm)
214
215	85x	if (nrow(l_df[[1]]) > 0 && nrow(l_df[[2]]) > 0) {
216		# Hazard ratio and CI.
217	79x	result <- s_coxph_pairwise(
218	79x	df = l_df[[2]],
219	79x	.ref_group = l_df[[1]],
220	79x	.in_ref_col = FALSE,
221	79x	.var = "tte",
222	79x	is_event = "is_event",
223	79x	strata = strata_vars,
224	79x	control = control
225		)
226
227	79x	df <- data.frame(
228		# Dummy column needed downstream to create a nested header.
229	79x	arm = " ",
230	79x	n_tot = unname(as.numeric(result$n_tot)),
231	79x	n_tot_events = unname(as.numeric(result$n_tot_events)),
232	79x	hr = unname(as.numeric(result$hr)),
233	79x	lcl = unname(result$hr_ci[1]),
234	79x	ucl = unname(result$hr_ci[2]),
235	79x	conf_level = control[["conf_level"]],
236	79x	pval = as.numeric(result$pvalue),
237	79x	pval_label = obj_label(result$pvalue),
238	79x	stringsAsFactors = FALSE
239		)
240		} else if (
241	6x	(nrow(l_df[[1]]) == 0 && nrow(l_df[[2]]) > 0) \|\|
242	6x	(nrow(l_df[[1]]) > 0 && nrow(l_df[[2]]) == 0)
243		) {
244	6x	df_tte_complete <- df_tte[stats::complete.cases(df_tte), ]
245	6x	df <- data.frame(
246		# Dummy column needed downstream to create a nested header.
247	6x	arm = " ",
248	6x	n_tot = nrow(df_tte_complete),
249	6x	n_tot_events = sum(df_tte_complete$is_event),
250	6x	hr = NA,
251	6x	lcl = NA,
252	6x	ucl = NA,
253	6x	conf_level = control[["conf_level"]],
254	6x	pval = NA,
255	6x	pval_label = NA,
256	6x	stringsAsFactors = FALSE
257		)
258		} else {
259	!	df <- data.frame(
260		# Dummy column needed downstream to create a nested header.
261	!	arm = " ",
262	!	n_tot = 0L,
263	!	n_tot_events = 0L,
264	!	hr = NA,
265	!	lcl = NA,
266	!	ucl = NA,
267	!	conf_level = control[["conf_level"]],
268	!	pval = NA,
269	!	pval_label = NA,
270	!	stringsAsFactors = FALSE
271		)
272		}
273
274	85x	df
275		}
276
277		#' @describeIn h_survival_duration_subgroups Summarizes estimates of the treatment hazard ratio
278		#' across subgroups in a data frame. `variables` corresponds to the names of variables found in
279		#' `data`, passed as a named list and requires elements `tte`, `is_event`, `arm` and
280		#' optionally `subgroups` and `strata`. `groups_lists` optionally specifies
281		#' groupings for `subgroups` variables.
282		#'
283		#' @return
284		#' * `h_coxph_subgroups_df()` returns a `data.frame` with columns `arm`, `n_tot`, `n_tot_events`, `hr`,
285		#' `lcl`, `ucl`, `conf_level`, `pval`, `pval_label`, `subgroup`, `var`, `var_label`, and `row_type`.
286		#'
287		#' @examples
288		#' # Extract hazard ratio for multiple groups.
289		#' h_coxph_subgroups_df(
290		#' variables = list(
291		#' tte = "AVAL",
292		#' is_event = "is_event",
293		#' arm = "ARM",
294		#' subgroups = c("SEX", "BMRKR2")
295		#' ),
296		#' data = adtte_f
297		#' )
298		#'
299		#' # Define groupings of BMRKR2 levels.
300		#' h_coxph_subgroups_df(
301		#' variables = list(
302		#' tte = "AVAL",
303		#' is_event = "is_event",
304		#' arm = "ARM",
305		#' subgroups = c("SEX", "BMRKR2")
306		#' ),
307		#' data = adtte_f,
308		#' groups_lists = list(
309		#' BMRKR2 = list(
310		#' "low" = "LOW",
311		#' "low/medium" = c("LOW", "MEDIUM"),
312		#' "low/medium/high" = c("LOW", "MEDIUM", "HIGH")
313		#' )
314		#' )
315		#' )
316		#'
317		#' # Extract hazard ratio for multiple groups with stratification factors.
318		#' h_coxph_subgroups_df(
319		#' variables = list(
320		#' tte = "AVAL",
321		#' is_event = "is_event",
322		#' arm = "ARM",
323		#' subgroups = c("SEX", "BMRKR2"),
324		#' strata = c("STRATA1", "STRATA2")
325		#' ),
326		#' data = adtte_f
327		#' )
328		#'
329		#' @export
330		h_coxph_subgroups_df <- function(variables,
331		data,
332		groups_lists = list(),
333		control = control_coxph(),
334		label_all = "All Patients") {
335	17x	if ("strat" %in% names(variables)) {
336	!	warning(
337	!	"Warning: the `strat` element name of the `variables` list argument to `h_coxph_subgroups_df() ",
338	!	"was deprecated in tern 0.9.4.\n ",
339	!	"Please use the name `strata` instead of `strat` in the `variables` argument."
340		)
341	!	variables[["strata"]] <- variables[["strat"]]
342		}
343
344	17x	checkmate::assert_character(variables$tte)
345	17x	checkmate::assert_character(variables$is_event)
346	17x	checkmate::assert_character(variables$arm)
347	17x	checkmate::assert_character(variables$subgroups, null.ok = TRUE)
348	17x	checkmate::assert_character(variables$strata, null.ok = TRUE)
349	17x	assert_df_with_factors(data, list(val = variables$arm), min.levels = 2, max.levels = 2)
350	17x	assert_df_with_variables(data, variables)
351	17x	checkmate::assert_string(label_all)
352
353		# Add All Patients.
354	17x	result_all <- h_coxph_df(
355	17x	tte = data[[variables$tte]],
356	17x	is_event = data[[variables$is_event]],
357	17x	arm = data[[variables$arm]],
358	17x	strata_data = if (is.null(variables$strata)) NULL else data[variables$strata],
359	17x	control = control
360		)
361	17x	result_all$subgroup <- label_all
362	17x	result_all$var <- "ALL"
363	17x	result_all$var_label <- label_all
364	17x	result_all$row_type <- "content"
365
366		# Add Subgroups.
367	17x	if (is.null(variables$subgroups)) {
368	3x	result_all
369		} else {
370	14x	l_data <- h_split_by_subgroups(data, variables$subgroups, groups_lists = groups_lists)
371
372	14x	l_result <- lapply(l_data, function(grp) {
373	64x	result <- h_coxph_df(
374	64x	tte = grp$df[[variables$tte]],
375	64x	is_event = grp$df[[variables$is_event]],
376	64x	arm = grp$df[[variables$arm]],
377	64x	strata_data = if (is.null(variables$strata)) NULL else grp$df[variables$strata],
378	64x	control = control
379		)
380	64x	result_labels <- grp$df_labels[rep(1, times = nrow(result)), ]
381	64x	cbind(result, result_labels)
382		})
383
384	14x	result_subgroups <- do.call(rbind, args = c(l_result, make.row.names = FALSE))
385	14x	result_subgroups$row_type <- "analysis"
386
387	14x	rbind(
388	14x	result_all,
389	14x	result_subgroups
390		)
391		}
392		}
393
394		#' Split data frame by subgroups
395		#'
396		#' @description `r lifecycle::badge("stable")`
397		#'
398		#' Split a data frame into a non-nested list of subsets.
399		#'
400		#' @inheritParams argument_convention
401		#' @inheritParams survival_duration_subgroups
402		#' @param data (`data.frame`)\cr dataset to split.
403		#' @param subgroups (`character`)\cr names of factor variables from `data` used to create subsets.
404		#' Unused levels not present in `data` are dropped. Note that the order in this vector
405		#' determines the order in the downstream table.
406		#'
407		#' @return A list with subset data (`df`) and metadata about the subset (`df_labels`).
408		#'
409		#' @details Main functionality is to prepare data for use in forest plot layouts.
410		#'
411		#' @examples
412		#' df <- data.frame(
413		#' x = c(1:5),
414		#' y = factor(c("A", "B", "A", "B", "A"), levels = c("A", "B", "C")),
415		#' z = factor(c("C", "C", "D", "D", "D"), levels = c("D", "C"))
416		#' )
417		#' formatters::var_labels(df) <- paste("label for", names(df))
418		#'
419		#' h_split_by_subgroups(
420		#' data = df,
421		#' subgroups = c("y", "z")
422		#' )
423		#'
424		#' h_split_by_subgroups(
425		#' data = df,
426		#' subgroups = c("y", "z"),
427		#' groups_lists = list(
428		#' y = list("AB" = c("A", "B"), "C" = "C")
429		#' )
430		#' )
431		#'
432		#' @export
433		h_split_by_subgroups <- function(data,
434		subgroups,
435		groups_lists = list()) {
436	66x	checkmate::assert_character(subgroups, min.len = 1, any.missing = FALSE)
437	66x	checkmate::assert_list(groups_lists, names = "named")
438	66x	checkmate::assert_subset(names(groups_lists), subgroups)
439	66x	assert_df_with_factors(data, as.list(stats::setNames(subgroups, subgroups)))
440
441	66x	data_labels <- unname(formatters::var_labels(data))
442	66x	df_subgroups <- data[, subgroups, drop = FALSE]
443	66x	subgroup_labels <- formatters::var_labels(df_subgroups, fill = TRUE)
444
445	66x	l_labels <- Map(function(grp_i, name_i) {
446	120x	existing_levels <- levels(droplevels(grp_i))
447	120x	grp_levels <- if (name_i %in% names(groups_lists)) {
448		# For this variable groupings are defined. We check which groups are contained in the data.
449	11x	group_list_i <- groups_lists[[name_i]]
450	11x	group_has_levels <- vapply(group_list_i, function(lvls) any(lvls %in% existing_levels), TRUE)
451	11x	names(which(group_has_levels))
452		} else {
453	109x	existing_levels
454		}
455	120x	df_labels <- data.frame(
456	120x	subgroup = grp_levels,
457	120x	var = name_i,
458	120x	var_label = unname(subgroup_labels[name_i]),
459	120x	stringsAsFactors = FALSE # Rationale is that subgroups may not be unique.
460		)
461	66x	}, df_subgroups, names(df_subgroups))
462
463		# Create a data frame with one row per subgroup.
464	66x	df_labels <- do.call(rbind, args = c(l_labels, make.row.names = FALSE))
465	66x	row_label <- paste0(df_labels$var, ".", df_labels$subgroup)
466	66x	row_split_var <- factor(row_label, levels = row_label)
467
468		# Create a list of data subsets.
469	66x	lapply(split(df_labels, row_split_var), function(row_i) {
470	294x	which_row <- if (row_i$var %in% names(groups_lists)) {
471	31x	data[[row_i$var]] %in% groups_lists[[row_i$var]][[row_i$subgroup]]
472		} else {
473	263x	data[[row_i$var]] == row_i$subgroup
474		}
475	294x	df <- data[which_row, ]
476	294x	rownames(df) <- NULL
477	294x	formatters::var_labels(df) <- data_labels
478
479	294x	list(
480	294x	df = df,
481	294x	df_labels = data.frame(row_i, row.names = NULL)
482		)
483		})
484		}

1		#' Survival time analysis
2		#'
3		#' @description `r lifecycle::badge("stable")`
4		#'
5		#' The analyze function [surv_time()] creates a layout element to analyze survival time by calculating survival time
6		#' median, median confidence interval, quantiles, and range (for all, censored, or event patients). The primary
7		#' analysis variable `vars` is the time variable and the secondary analysis variable `is_event` indicates whether or
8		#' not an event has occurred.
9		#'
10		#' @inheritParams argument_convention
11		#' @param control (`list`)\cr parameters for comparison details, specified by using the helper function
12		#' [control_surv_time()]. Some possible parameter options are:
13		#' * `conf_level` (`proportion`)\cr confidence level of the interval for survival time.
14		#' * `conf_type` (`string`)\cr confidence interval type. Options are "plain" (default), "log", or "log-log",
15		#' see more in [survival::survfit()]. Note option "none" is not supported.
16		#' * `quantiles` (`numeric`)\cr vector of length two to specify the quantiles of survival time.
17		#' @param ref_fn_censor (`flag`)\cr whether referential footnotes indicating censored observations should be printed
18		#' when the `range` statistic is included.
19		#' @param .stats (`character`)\cr statistics to select for the table. Run `get_stats("surv_time")`
20		#' to see available statistics for this function.
21		#' @param .indent_mods (named `integer`)\cr indent modifiers for the labels. Each element of the vector
22		#' should be a name-value pair with name corresponding to a statistic specified in `.stats` and value the indentation
23		#' for that statistic's row label.
24		#'
25		#' @examples
26		#' library(dplyr)
27		#'
28		#' adtte_f <- tern_ex_adtte %>%
29		#' filter(PARAMCD == "OS") %>%
30		#' mutate(
31		#' AVAL = day2month(AVAL),
32		#' is_event = CNSR == 0
33		#' )
34		#' df <- adtte_f %>% filter(ARMCD == "ARM A")
35		#'
36		#' @name survival_time
37		#' @order 1
38		NULL
39
40		#' @describeIn survival_time Statistics function which analyzes survival times.
41		#'
42		#' @return
43		#' * `s_surv_time()` returns the statistics:
44		#' * `median`: Median survival time.
45		#' * `median_ci`: Confidence interval for median time.
46		#' * `quantiles`: Survival time for two specified quantiles.
47		#' * `range_censor`: Survival time range for censored observations.
48		#' * `range_event`: Survival time range for observations with events.
49		#' * `range`: Survival time range for all observations.
50		#'
51		#' @keywords internal
52		s_surv_time <- function(df,
53		.var,
54		is_event,
55		control = control_surv_time()) {
56	232x	checkmate::assert_string(.var)
57	232x	assert_df_with_variables(df, list(tte = .var, is_event = is_event))
58	232x	checkmate::assert_numeric(df[[.var]], min.len = 1, any.missing = FALSE)
59	232x	checkmate::assert_logical(df[[is_event]], min.len = 1, any.missing = FALSE)
60
61	232x	conf_type <- control$conf_type
62	232x	conf_level <- control$conf_level
63	232x	quantiles <- control$quantiles
64
65	232x	formula <- stats::as.formula(paste0("survival::Surv(", .var, ", ", is_event, ") ~ 1"))
66	232x	srv_fit <- survival::survfit(
67	232x	formula = formula,
68	232x	data = df,
69	232x	conf.int = conf_level,
70	232x	conf.type = conf_type
71		)
72	232x	srv_tab <- summary(srv_fit, extend = TRUE)$table
73	232x	srv_qt_tab <- stats::quantile(srv_fit, probs = quantiles)$quantile
74	232x	range_censor <- range_noinf(df[[.var]][!df[[is_event]]], na.rm = TRUE)
75	232x	range_event <- range_noinf(df[[.var]][df[[is_event]]], na.rm = TRUE)
76	232x	range <- range_noinf(df[[.var]], na.rm = TRUE)
77	232x	list(
78	232x	median = formatters::with_label(unname(srv_tab["median"]), "Median"),
79	232x	median_ci = formatters::with_label(
80	232x	unname(srv_tab[paste0(srv_fit$conf.int, c("LCL", "UCL"))]), f_conf_level(conf_level)
81		),
82	232x	quantiles = formatters::with_label(
83	232x	unname(srv_qt_tab), paste0(quantiles[1] * 100, "% and ", quantiles[2] * 100, "%-ile")
84		),
85	232x	range_censor = formatters::with_label(range_censor, "Range (censored)"),
86	232x	range_event = formatters::with_label(range_event, "Range (event)"),
87	232x	range = formatters::with_label(range, "Range")
88		)
89		}
90
91		#' @describeIn survival_time Formatted analysis function which is used as `afun` in `surv_time()`.
92		#'
93		#' @return
94		#' * `a_surv_time()` returns the corresponding list with formatted [rtables::CellValue()].
95		#'
96		#' @examples
97		#' a_surv_time(
98		#' df,
99		#' .df_row = df,
100		#' .var = "AVAL",
101		#' is_event = "is_event"
102		#' )
103		#'
104		#' @export
105		a_surv_time <- function(df,
106		labelstr = "",
107		.var = NULL,
108		.df_row = NULL,
109		is_event,
110		control = control_surv_time(),
111		ref_fn_censor = TRUE,
112		.stats = NULL,
113		.formats = NULL,
114		.labels = NULL,
115		.indent_mods = NULL,
116		na_str = default_na_str()) {
117	14x	x_stats <- s_surv_time(
118	14x	df = df, .var = .var, is_event = is_event, control = control
119		)
120	14x	rng_censor_lwr <- x_stats[["range_censor"]][1]
121	14x	rng_censor_upr <- x_stats[["range_censor"]][2]
122
123		# Use method-specific defaults
124	14x	fmts <- c(median_ci = "(xx.x, xx.x)", quantiles = "xx.x, xx.x", range = "xx.x to xx.x")
125	14x	lbls <- c(median_ci = "95% CI", range = "Range", range_censor = "Range (censored)", range_event = "Range (event)")
126	14x	lbls_custom <- .labels
127	14x	.formats <- c(.formats, fmts[setdiff(names(fmts), names(.formats))])
128	14x	.labels <- c(.labels, lbls[setdiff(names(lbls), names(lbls_custom))])
129
130		# Fill in with formatting defaults if needed
131	14x	.stats <- get_stats("surv_time", stats_in = .stats)
132	14x	.formats <- get_formats_from_stats(.stats, .formats)
133	14x	.labels <- get_labels_from_stats(.stats, .labels) %>% labels_use_control(control, lbls_custom)
134	14x	.indent_mods <- get_indents_from_stats(.stats, .indent_mods)
135
136	14x	x_stats <- x_stats[.stats]
137
138		# Auto format handling
139	14x	.formats <- apply_auto_formatting(.formats, x_stats, .df_row, .var)
140
141	14x	cell_fns <- setNames(vector("list", length = length(x_stats)), .labels)
142	14x	if ("range" %in% names(x_stats) && ref_fn_censor) {
143	14x	if (identical(x_stats[["range"]][1], rng_censor_lwr) && identical(x_stats[["range"]][2], rng_censor_upr)) {
144	2x	cell_fns[[.labels[["range"]]]] <- "Censored observations: range minimum & maximum"
145	12x	} else if (identical(x_stats[["range"]][1], rng_censor_lwr)) {
146	2x	cell_fns[[.labels[["range"]]]] <- "Censored observation: range minimum"
147	10x	} else if (identical(x_stats[["range"]][2], rng_censor_upr)) {
148	1x	cell_fns[[.labels[["range"]]]] <- "Censored observation: range maximum"
149		}
150		}
151
152	14x	in_rows(
153	14x	.list = x_stats,
154	14x	.formats = .formats,
155	14x	.names = .labels,
156	14x	.labels = .labels,
157	14x	.indent_mods = .indent_mods,
158	14x	.format_na_strs = na_str,
159	14x	.cell_footnotes = cell_fns
160		)
161		}
162
163		#' @describeIn survival_time Layout-creating function which can take statistics function arguments
164		#' and additional format arguments. This function is a wrapper for [rtables::analyze()].
165		#'
166		#' @return
167		#' * `surv_time()` returns a layout object suitable for passing to further layouting functions,
168		#' or to [rtables::build_table()]. Adding this function to an `rtable` layout will add formatted rows containing
169		#' the statistics from `s_surv_time()` to the table layout.
170		#'
171		#' @examples
172		#' basic_table() %>%
173		#' split_cols_by(var = "ARMCD") %>%
174		#' add_colcounts() %>%
175		#' surv_time(
176		#' vars = "AVAL",
177		#' var_labels = "Survival Time (Months)",
178		#' is_event = "is_event",
179		#' control = control_surv_time(conf_level = 0.9, conf_type = "log-log")
180		#' ) %>%
181		#' build_table(df = adtte_f)
182		#'
183		#' @export
184		#' @order 2
185		surv_time <- function(lyt,
186		vars,
187		is_event,
188		control = control_surv_time(),
189		ref_fn_censor = TRUE,
190		na_str = default_na_str(),
191		nested = TRUE,
192		...,
193		var_labels = "Time to Event",
194		show_labels = "visible",
195		table_names = vars,
196		.stats = c("median", "median_ci", "quantiles", "range"),
197		.formats = NULL,
198		.labels = NULL,
199		.indent_mods = c(median_ci = 1L)) {
200	3x	extra_args <- list(
201	3x	.stats = .stats, .formats = .formats, .labels = .labels, .indent_mods = .indent_mods, na_str = na_str,
202	3x	is_event = is_event, control = control, ref_fn_censor = ref_fn_censor, ...
203		)
204
205	3x	analyze(
206	3x	lyt = lyt,
207	3x	vars = vars,
208	3x	afun = a_surv_time,
209	3x	var_labels = var_labels,
210	3x	show_labels = show_labels,
211	3x	table_names = table_names,
212	3x	na_str = na_str,
213	3x	nested = nested,
214	3x	extra_args = extra_args
215		)
216		}