Lost row after rbind or subsetting?

[Melkiades]

I was trying to implement the nested = FALSE logic to make the rbind work without having to subset (as discussed yesterday) but the "Overall" row disappears. So I tried to make a simplified version that uses only rtables functions and here it is (my apologies for the mean function being a bit weird):

library(rtables)
lyt <- basic_table() %>%
  split_rows_by(
    var = "ARM",
    label_pos = "topleft",
    split_label = "Treatment Group",
    split_fun = drop_split_levels,
    section_div = ""
  ) %>%
  add_rowcounts() %>%
  split_rows_by(
    var = "STRATA1",
    label_pos = "topleft",
    split_label = "Visit"
  )

# I know this is ugly but I manage like this
my_mean <- function(x, ...) list("a" = mean(x, ...))
afun_list <- list("mean 1" = make_afun(my_mean, .stats = "a", .labels = "Overall"),
                  "mean 2" = make_afun(my_mean, .stats = "a",  .labels = "Overall"))

lyt_std <- lyt %>%
  split_cols_by_multivar(
    vars = c("BMRKR1", "AGE"), 
    varlabels = c("mean 1", "mean 2")
  ) %>% 
  analyze_colvars(list(mean, mean), format = "xx.x")

lyt_overall <- basic_table() %>%
  split_cols_by_multivar(
    vars = c("BMRKR1", "AGE"), 
    varlabels = c("mean 1", "mean 2")
  ) %>% 
  analyze_colvars(afun_list, 
                  format = "xx.x", 
                  nested = TRUE, 
                  extra_args = list(".labels" = "Overall")
                  )
  

result_tbl <- build_table(lyt_std, DM)
result_overall <- build_table(lyt_overall, DM)
result_overall # This produces the right row (1)
col_info(result_tbl) <- col_info(result_overall)
result <- rbind(result_tbl, result_overall)
result # row (1) disappears
table_structure(result)

Ideally, we would like to get rid of the rbind entirely and at the moment we managed only like this:

tbl <- basic_table() %>%
  split_rows_by(
    var = "ARM",
    label_pos = "topleft",
    split_label = "Treatment Group",
    split_fun = add_overall_level(
      trim = TRUE, first = FALSE,
      valname = "ARM_ALL", label = "Overall"
    ),
    section_div = ""
  ) %>%
  add_rowcounts() %>%
  split_rows_by(
    var = "STRATA1",
    label_pos = "topleft",
    split_label = "Strata 1",
    split_fun = add_overall_level(
      trim = TRUE, first = FALSE,
      valname = "strata_ALL", label = "Overall"
    )
  ) %>%
  split_cols_by_multivar(
    vars = c("BMRKR1", "AGE"),
    varlabels = c("mean 1", "mean 2")
  ) %>%
  analyze_colvars(list(mean, mean), format = "xx.x") %>%
  build_table(DM)
tbl[ -c(c(8, 9), c(16, 17), c(26, 27), 29:35), ] # horrible but we get the right table

[gmbecker]

This is what you want, although it is surfacing a bug, as I'm not sure why the last row's row label is not being printed:

tbl <- basic_table() %>%
  split_rows_by(
    var = "ARM",
    label_pos = "topleft",
    split_label = "Treatment Group",
    section_div = ""
  ) %>%
  add_rowcounts() %>%
  split_rows_by(
    var = "STRATA1",
    label_pos = "topleft",
    split_label = "Strata 1",
    split_fun = add_overall_level(
      trim = TRUE, first = FALSE,
      valname = "strata_ALL", label = "Overall"
    )
  ) %>%
  split_cols_by_multivar(
    vars = c("BMRKR1", "AGE"),
    varlabels = c("mean 1", "mean 2")
  ) %>%
    analyze_colvars(list(mean, mean), format = "xx.x") %>%
    analyze_colvars(afun_list,
                  format = "xx.x",
                  nested = FALSE,
                  extra_args = list(".labels" = "Overall")
                  ) %>%
  build_table(DM)

[Melkiades]

could it be due to the fact that the second table (overall) is an elementaryTable instead of a tableTree?

[gmbecker]

This is caused by the above bug in formatters, is fixed by insightsengineering/formatters#78

After that PR is merged in, and provided you have formatters >= 0.3.4.11, you will get:

> tbl <- basic_table() %>%
+   split_rows_by(
+     var = "ARM",
+     label_pos = "topleft",
+     split_label = "Treatment Group",
+     section_div = ""
+   ) %>%
+   add_rowcounts() %>%
+   split_rows_by(
+     var = "STRATA1",
+     label_pos = "topleft",
+     split_label = "Strata 1",
+     split_fun = add_overall_level(
+       trim = TRUE, first = FALSE,
+       valname = "strata_ALL", label = "Overall"
+     )
+   ) %>%
+   split_cols_by_multivar(
+     vars = c("BMRKR1", "AGE"),
+     varlabels = c("mean 1", "mean 2")
+   ) %>%
+     analyze_colvars(list(mean, mean), format = "xx.x") %>%
+     analyze_colvars(afun_list,
+                   format = "xx.x",
+                   nested = FALSE,
+                   extra_args = list(".labels" = "Overall")
+                   ) %>%
+   build_table(DM)
> tbl
Treatment Group                         
  Strata 1               mean 1   mean 2
————————————————————————————————————————
A: Drug X (N=121)                       
  A                                     
                          5.9      32.5 
  B                                     
                          5.5      35.5 
  C                                     
                          5.9      36.3 
  Overall                               
                          5.8      34.9 

B: Placebo (N=106)                      
  A                                     
                          6.0      32.3 
  B                                     
                          5.9      32.4 
  C                                     
                          6.5      34.5 
  Overall                               
                          6.1      33.0 

C: Combination (N=129)                  
  A                                     
                          6.0      35.8 
  B                                     
                          6.3      34.4 
  C                                     
                          4.9      33.5 
  Overall                               
                          5.7      34.6 

Overall                   5.9      34.2 

As desried. Note, if you don't want the levels of STRATA1 and their corresponding mean values to be on different lines, you would remove the inner split_rows_by and have analysis functions which (all) return length(levels(STRATA1)) + 1 rows via in_rows.

[Melkiades]

Thanks a lot! It works well