Add warning for missing p-value in tabulate_rsp_subgroups() output (#…

…1296)

Fixes #1289

NEWS.md

@@ -4,6 +4,7 @@
 * Reworking of `summarize_glm_count()` documentation and all its associated functions to better describe the results and the functions' purpose.
 * Added the `.formats` argument to `tabulate_rsp_subgroups` and `tabulate_survival_subgroups` to allow users to specify formats.
 * Added the `riskdiff` argument to `tabulate_rsp_subgroups` and `tabulate_survival_subgroups` to allow users to add a risk difference table column, and function `control_riskdiff` to specify settings for the risk difference column.
+* Added warning to `tabulate_rsp_subgroups` when `pval` statistic is selected but `df` has not been correctly generated to add p-values to the output table.
 
 ### Bug Fixes
 * Fixed a bug in `a_surv_time` that threw an error when split only has `"is_event"`.
@@ -15,7 +16,6 @@
 * Fixed bug in `decorate_grob` that did not handle well empty strings or `NULL` values for title and footers.
 * Fixed bug in `g_km` that caused an error when multiple records in the data had estimates at max time.
 
-
 ### Miscellaneous
 * Began deprecation of the confusing functions `summary_formats` and `summary_labels`.
 

R/argument_convention.R

@@ -43,7 +43,7 @@
 #'   for more information.
 #' @param lyt (`PreDataTableLayouts`)\cr layout that analyses will be added to.
 #' @param method (`string` or `NULL`)\cr specifies the test used to calculate the p-value for the difference between
-#'   two proportions. For options, see [s_test_proportion_diff()]. Default is `NULL` so no test is performed.
+#'   two proportions. For options, see [test_proportion_diff()]. Default is `NULL` so no test is performed.
 #' @param na.rm (`flag`)\cr whether `NA` values should be removed from `x` prior to analysis.
 #' @param na_str (`string`)\cr string used to replace all `NA` or empty values in the output.
 #' @param nested (`flag`)\cr whether this layout instruction should be applied within the existing layout structure _if

R/response_subgroups.R

@@ -220,6 +220,14 @@ tabulate_rsp_subgroups <- function(lyt,
                                    )) {
   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)

tests/testthat/test-response_subgroups.R

@@ -307,3 +307,24 @@ testthat::test_that("tabulate_rsp_subgroups riskdiff argument works as expected"
   res <- testthat::expect_silent(result)
   testthat::expect_snapshot(res)
 })
+
+testthat::test_that("tabulate_rsp_subgroups pval statistic warning works as expected", {
+  adrs <- adrs_200
+
+  df <- extract_rsp_subgroups(
+    variables = list(rsp = "rsp", arm = "ARM", subgroups = c("SEX", "STRATA2")),
+    data = adrs,
+    method = NULL,
+    conf_level = 0.95
+  )
+
+  # warning when no pval in df
+  expect_warning(
+    basic_table() %>%
+      tabulate_rsp_subgroups(
+        df = df,
+        vars = c("n", "prop", "n_tot", "or", "ci", "pval")
+      ),
+    "please specify a p-value test"
+  )
+})