Variable grouping in gtsummary with dichotomous variable

I am trying to represent multiple dichotomous variables as a subgroup of another variable.

Below a sample of my current code:

set.seed(123)
library(tidyverse)
library(gtsummary)
library(bstfun)

d <- tidyr::tibble(
  id = c(1:20),
  et_yn = structure(factor(sample(seq(0,1), size = 20, replace = TRUE), levels = c(0,1), labels = c("No", "Yes")), label = "ET")
) %>%
  rowwise() %>%
  mutate(
    et_typ_1 = factor(ifelse(et_yn == "Yes", sample(seq(0,1), size = 1, replace = T), NA), levels = c(0,1), labels = c("No", "Yes")),
    et_typ_2 = factor(ifelse(et_yn == "Yes", sample(seq(0,1), size = 1, replace = T), NA), levels = c(0:1), labels = c("No", "Yes")),
    et_gp_1 = factor(ifelse(et_typ_2 == "Yes", sample(seq(0,1), size = 1, replace = T), NA), levels = c(0,1), labels = c("No", "Yes")),
    et_gp_2 = factor(ifelse(et_typ_2 == "Yes", sample(seq(0,1), size = 1, replace = T), NA), levels = c(0,1), labels = c("No", "Yes")),
    et_gp_3 = factor(ifelse(et_typ_2 == "Yes", sample(seq(0,1), size = 1, replace = T), NA), levels = c(0,1), labels = c("No", "Yes")),
  ) %>%
  data.frame() %>%
  mutate(
    et_typ_2 = structure(et_typ_2, label = "ET gp"),
    et_gp_1 = structure(et_gp_1, label = "St.a"),
    et_gp_2 = structure(et_gp_2, label = "St.b"),
    et_gp_3 = structure(et_gp_3, label = "St.c"),
  )

d %>%
  tbl_summary(
    data = .,
    include = -id,
    missing = "no",
  ) %>%
  add_n() %>%
  modify_header(stat_0 = "**n (%)**") %>%
  bstfun::add_variable_grouping(
    "GP" = c("et_gp_1", "et_gp_2", "et_gp_3"),
  ) %>%
  gtsummary::as_kable()

^{Created on 2024-12-30 with reprex v2.1.1}

However the output I would like is to have all St.<x> without an N and the GP row with that same N, like so:

I tried appending the following code at the bottom of the previous code:

d %>%
  tbl_summary(
    data = .,
    include = -id,
    missing = "no",
  ) %>%
  add_n() %>%
  modify_header(stat_0 = "**n (%)**") %>%
  bstfun::add_variable_grouping(
    "GP" = c("et_gp_1", "et_gp_2", "et_gp_3"),
  ) %>%
  modify_table_body(
    fun = ~ .x %>%
      mutate(n = ifelse(variable == "GP", max(n), NA))
  )

which doesn't unfortunately result in the desired output and actually seems to have no effect.

What am I doing wrong?

The bstfun::add_variable_grouping() function is also a piece of code I would like to replace.

Thank you.

Here's an example of getting the table you're after.

1. Build your basic table
2. Indent the grouping variables
3. Add a row for the grouping header
4. Remove the Ns for the grouping variables, and add the N to the group header level.

library(gtsummary)

tbl <- trial |> 
  tbl_summary(
    include = c("age", "response", "death"),
    type = c("response", "death") ~ "dichotomous",
    missing = "no"
  ) |> 
  add_n()

# print the table_body to see what is available to manipulate
tbl$table_body
#> # A tibble: 3 × 7
#>   variable var_type    row_type var_label      label          n     stat_0     
#>   <chr>    <chr>       <chr>    <chr>          <chr>          <chr> <chr>      
#> 1 age      continuous  label    Age            Age            189   47 (38, 57)
#> 2 response dichotomous label    Tumor Response Tumor Response 193   61 (32%)   
#> 3 death    dichotomous label    Patient Died   Patient Died   200   112 (56%)

# make final modificatons to the table
tbl |> 
  # begin by indenting the sub variables
  modify_column_indent(
    columns = "label",
    rows = .data$variable %in% c("response", "death"),
    indent = 4
  ) |> 
  # next add a row for the grouping
  modify_table_body(
    ~ .x |> 
      tibble::add_row(label = "Treatment Outcome", .before = 2L)
  ) |> 
  # finally, remove Ns for the subgroups, and add an N to the grouping level
  modify_table_body(
    ~ .x |> 
      dplyr::mutate(
        n = dplyr::case_when(
          # put any number you want here, it's not always the case that the subgroups 
          # will have the same number of non-missing observations (as in this example)
          label == "Treatment Outcome" ~ stringr::str_glue("{nrow(trial)}"), 
          variable %in% c("response", "death") ~ NA,
          .default = .data$n
        )
      )
  ) |> 
  as_kable() # convert to kable to display on stackoverflow

^{Created on 2024-12-30 with reprex v2.1.1}