Using rbind to combine two tables when add_combo_levels is needed

[yutong1307]

Hello,

I am trying to use rbind to combine two tables generated seperately. The following example works when there are only treatment columns in the tables I want to combine:

library(rtables)

l1 <- basic_table() %>%
  split_cols_by("ARM") %>%
  add_colcounts() %>%
  analyze("AGE")

tab1 <- build_table(l1, DM)

l2 <- basic_table() %>%
  split_cols_by("ARM") %>%
  add_colcounts() %>%
  analyze("SEX")

tab2 <- build_table(l2, DM)
rbind(tab1, tab2)

However, when I tried to include columns with combined count from two different treatment arms using add_combo_levels, I would received an error "Error in FUN(X[[i]], ...) : attempted to add child with non-matching, non-empty column info to an existing table"

library(rtables)
library(tibble)

combodf <- tribble(
  ~valname, ~label, ~levelcombo, ~exargs,
  "A_B", "Arms A+B", c("A: Drug X", "B: Placebo"), list(),
  "A_C", "Arms A+C", c("A: Drug X", "C: Combination"), list())

l1 <- basic_table() %>%
  split_cols_by("ARM", split_fun = add_combo_levels(combodf)) %>%
  add_colcounts() %>%
  analyze("AGE")

tab1 <- build_table(l1, DM)

l2 <- basic_table() %>%
  split_cols_by("ARM", split_fun = add_combo_levels(combodf)) %>%
  add_colcounts() %>%
  analyze("SEX")

tab2 <- build_table(l2, DM)
rbind(tab1, tab2)

Is there anyway to solve this problem?

[gmbecker]

Hi @yutong1307 at first glance this looks like its simply a bug. I don't see any reasons that the column infos of the two tables should be found to not match.

In the mean time, the work around to get a "sloppy" rbind where only very minimal checking for conformity is done is to set the col_info of the second table to EmptyColInfo which bypasses the check (which is generally important but appears to have a bug here):

library(rtables)
library(tibble)

combodf <- tribble(
  ~valname, ~label, ~levelcombo, ~exargs,
  "A_B", "Arms A+B", c("A: Drug X", "B: Placebo"), list(),
  "A_C", "Arms A+C", c("A: Drug X", "C: Combination"), list())

l1 <- basic_table() %>%
  split_cols_by("ARM", split_fun = add_combo_levels(combodf)) %>%
  add_colcounts() %>%
  analyze("AGE")

tab1 <- build_table(l1, DM)

l2 <- basic_table() %>%
  split_cols_by("ARM", split_fun = add_combo_levels(combodf)) %>%
  add_colcounts() %>%
  analyze("SEX")

tab2 <- build_table(l2, DM)

col_info(tab2) <- EmptyColInfo

tab3 <- rbind(tab1, tab2)

Which successfully gives us

> tab3
                   A: Drug X   B: Placebo   C: Combination   Arms A+B   Arms A+C
                    (N=121)     (N=106)        (N=129)       (N=227)    (N=250) 
————————————————————————————————————————————————————————————————————————————————
Mean                 34.91       33.02          34.57         34.03      34.73  
F                     70           56             61           126        131   
M                     51           50             68           101        119   
U                      0           0              0             0          0    
UNDIFFERENTIATED       0           0              0             0          0    

(which, despite its strange look, is correct in for your reprex).

[gmbecker]

This was a bug that is now fixed, but if you need the behavior in the existing release, please use the workaround mentioned above.

The expected behavior will work without the workaround in the next release of rtables

Thanks for the report.