Empty levels missing when reference column is set

[gmbecker]

Hi @gmbecker, I think it is an issue with VarLevWBaselineSplit, i.e.

library(rtables)


df <- data.frame(
  val = 1:10,
  grp = factor(rep("a", 10), levels = c("a", "b"))
)

And

basic_table() |>
  split_cols_by("grp") |>
  analyze("val") |>
  build_table(df)

results in

        a     b 
————————————————
Mean   5.50   NA

and

basic_table() |>
  split_cols_by("grp", ref_group = "a") |>
  analyze("val") |>
  build_table(df)

results in

        a  
———————————
Mean   5.50

Can you please check that the level with no observation does not get dropped when setting the ref_group?

Originally posted by @waddella in #322 (comment)

[gmbecker]

So this is interesting. The code is very old (and, I think, not hit very often), but looking now at what is in there, this was at one point intentional. droplevels is being called explicitly when the global child-value-order is determined in the split-that-has-a-reference-group case.

I think that its ok to remove this. That already happened in the no reference-group case, from the looks of it. In both cases, teh columns are degenerate, but having a reference group doesn't make that more wrong in any way that i can think of. Note, however, that this is technically a user-facing behavior change, rather than a bugfix.

I'll bring it up in the meeting tomorrow and go ahead and make the change provided there's no pushback.

@danielinteractive fyi

[danielinteractive]

@gmbecker how did the meeting go? can you make the change?

[gmbecker]

Yes we will be making the change and it will be in the CRAN release that will be going out in the next couple days.