[Question]: custom summarize in group_desc () to obtain total n under same subgroup

[haodafa2019]

Description

I tried to create table for frequency and descriptive summary for variable SEX and AGE.
For SEX, I can obtain different pct by define set_denoms_by() which is very user friendly.
But for AAGE, I can obtain non-missing n and missing by each subgroup (TRT01A, AAGEGR1), but I could not obtain the denominator by group (TRT01A) or (TRT01AN, ARACE), do you have suggestion?

Code snippet

demo <- tplyr_table(adsl, TRT01A, cols = AAGEGR1) %>% 
  set_pop_data(adsl) %>% 
  set_pop_treat_var(TRT01A) %>% 
  add_total_group(group_name = "99") %>%
  add_layer(
    group_count(SEX,by=ARACE) %>%
      add_total_row(f_str(""), count_missings=FALSE, sort_value=-999) %>%
      set_format_strings(f_str("xx (x.x)", n, pct))%>%
      set_missing_count(f_str("xx (x.x)", n, pct), Missing = NA, sort_value = 98) %>% 
      set_total_row_label("Sex, n (%)") %>% 
      set_denoms_by(TRT01A,AAGEGR1,ARACE)
   ) %>% 
  add_layer(
    group_desc(AAGE, by=ARACE) %>% 
      set_custom_summaries(
        blank_=as.numeric(NA),
        nmiss=sum(!is.na(.var)),
        nmiss_pct=sum(!is.na(.var))/n*100,
        missing = sum(is.na(.var)),
        missing_pct= sum(is.na(.var))/n*100
      ) %>% 
      set_format_strings(
        "Age (Years)"  = f_str("xx", blank_),
        "Total n"            = f_str("xx",n),
        "n (%)"       = f_str("xx (x.x)", nmiss, nmiss_pct),
        "Missing (%)" = f_str("xx (x.x)", missing, missing_pct),
        "Mean +/-SD" = f_str("xx.x +/-x.xx", mean,sd, empty='NA'),
        "Median"    = f_str("xx.x", median),
        "Q1; Q3"    = f_str("xx.x; xx.x", q1, q3, empty=c(.overall='NA')),
        "Min; Max"  = f_str("xx; xx", min, max)
      ) 
  )

[mstackhouse]

Hi @haodafa2019. Interesting use case - unfortunately I can't think of a clean way to do this at the moment. I'll be honest - you taught me something new here because I didn't realize that n would be accessible in custom summaries...

Custom denominators are only calculated within a count or shift layer. A really hacky way to do this would be to create a count layer of AAGE and filter it out to only the missing counts, but that's very roundabout - so there's not a very smooth way to do it. Given that you can apparently accesss the other summaries, maybe we could attach denominators and bring the denoms by over to the desc layers...

I'm going to add this as another wishlist item for tplyr2 where we can handle it more robustly in the future as well.

[haodafa2019]

Hi @mstackhouse , aha, that is what I did. I use group_count() to obtain the n (%) and missing (%). and considering the big categories of continued variable, I would derive new category variable to calculate.

And regarding wish list, I use Tplyr a lot and now explore the use on multiple periods analysis in phase 1 studies, and I have a wishlist :) . I am very glad if we can have talk.