How to include second row split within the page split?

[danielinteractive]

Hi @gmbecker ,

I ran into this problem, where I would appreciate your help.

Starting point:

lyt <- basic_table() |> 
  split_cols_by("ARM") |> 
  split_rows_by("STRATA1", page_by = TRUE) |> 
  split_rows_by("RACE", split_fun = keep_split_levels("ASIAN")) |> 
  analyze("AGE", mean, var_labels = "Age", format = "xx.xx") |> 
  split_rows_by("RACE", split_fun = keep_split_levels("BLACK OR AFRICAN AMERICAN")) |> 
  analyze("AGE", mean, var_labels = "Age", format = "xx.xx") |> 
  analyze("BMRKR1", mean, var_labels = "Biomarker", format = "xx.xx")

build_table(lyt, DM)

Gives:

                            A: Drug X   B: Placebo   C: Combination
———————————————————————————————————————————————————————————————————
A                                                                  
  ASIAN                                                            
    mean                      32.19       33.90          36.81     
B                                                                  
  ASIAN                                                            
    mean                      34.12       31.62          34.73     
C                                                                  
  ASIAN                                                            
    mean                      36.21       33.00          32.39     
BLACK OR AFRICAN AMERICAN                                          
  Age                                                              
    mean                      34.68       31.71          34.00     
  Biomarker                                                        
    mean                      5.30         7.28           5.36 

But I would like to produce a table where the second row split of RACE is nested within the page split. So something like:

                            A: Drug X   B: Placebo   C: Combination
———————————————————————————————————————————————————————————————————
A                                                                  
  ASIAN                                                            
    mean                      32.19       33.90          36.81
  BLACK OR AFRICAN AMERICAN                                          
    Age                                                              
      mean                      34.68       31.71          34.00     
    Biomarker                                                        
      mean                      5.30         7.28           5.36      
B                                                                  
  ASIAN                                                            
    mean                      34.12       31.62          34.73     
  BLACK OR AFRICAN AMERICAN                                          
    Age                                                              
      mean                      34.68       31.71          34.00     
    Biomarker                                                        
      mean                      5.30         7.28           5.36 
C                                                                  
  ASIAN                                                            
    mean                      36.21       33.00          32.39     
  BLACK OR AFRICAN AMERICAN                                          
    Age                                                              
      mean                      34.68       31.71          34.00     
    Biomarker                                                        
      mean                      5.30         7.28           5.36 

How can I do that?

Edit: Actually it is not just for page by row split, it also is the same question for a normal row split:

lyt <- basic_table() |> 
  split_cols_by("ARM") |> 
  split_rows_by("STRATA1") |> 
  split_rows_by("RACE", split_fun = keep_split_levels("ASIAN")) |> 
  analyze("AGE", mean, var_labels = "Age", format = "xx.xx") |> 
  split_rows_by("RACE", split_fun = keep_split_levels("BLACK OR AFRICAN AMERICAN")) |> 
  analyze("AGE", mean, var_labels = "Age", format = "xx.xx") |> 
  analyze("BMRKR1", mean, var_labels = "Biomarker", format = "xx.xx")

build_table(lyt, DM)

gives:

                            A: Drug X   B: Placebo   C: Combination
———————————————————————————————————————————————————————————————————
A                                                                  
  ASIAN                                                            
    mean                      32.19       33.90          36.81     
B                                                                  
  ASIAN                                                            
    mean                      34.12       31.62          34.73     
C                                                                  
  ASIAN                                                            
    mean                      36.21       33.00          32.39     
BLACK OR AFRICAN AMERICAN                                          
  Age                                                              
    mean                      34.68       31.71          34.00     
  Biomarker                                                        
    mean                      5.30         7.28           5.36

> sessionInfo()
R version 4.4.2 (2024-10-31 ucrt)
Platform: x86_64-w64-mingw32/x64
Running under: Windows 11 x64 (build 26100)

Matrix products: default


locale:
[1] LC_COLLATE=German_Switzerland.utf8  LC_CTYPE=German_Switzerland.utf8   
[3] LC_MONETARY=German_Switzerland.utf8 LC_NUMERIC=C                       
[5] LC_TIME=German_Switzerland.utf8    

time zone: Asia/Taipei
tzcode source: internal

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] rtables_0.6.11    magrittr_2.0.3    formatters_0.5.10

loaded via a namespace (and not attached):
 [1] compiler_4.4.2    backports_1.5.0   fastmap_1.2.0     cli_3.6.3         htmltools_0.5.8.1
 [6] tools_4.4.2       rstudioapi_0.17.1 stringi_1.8.4     grid_4.4.2        checkmate_2.3.2  
[11] pkgload_1.4.0     digest_0.6.37     lifecycle_1.0.4   rlang_1.1.5   

Thanks,
cheers
Daniel

[danielinteractive]

OK it seems a combination of a split dependent analysis function and pruning the table can work:

afun <- function(x, .spl_context, exclude = "") {
  val <- .spl_context[nrow(.spl_context), "value"]
  if (val %in% exclude) {
    NULL
  } else {
    mean(x)
  }
}

lyt <- basic_table() |> 
  split_cols_by("ARM") |> 
  split_rows_by("STRATA1", page_by = TRUE) |> 
  split_rows_by("RACE", split_fun = keep_split_levels(c("ASIAN", "WHITE"))) |> 
  analyze("AGE", mean, var_labels = "Age", format = "xx.xx") |> 
  analyze("BMRKR1", afun, var_labels = "Biomarker", extra_args = list(exclude = "ASIAN"), format = "xx.xx")

tab <- build_table(lyt, DM)
prune_table(tab)

gives

                A: Drug X   B: Placebo   C: Combination
———————————————————————————————————————————————————————
A                                                      
  ASIAN                                                
    Age                                                
      mean        32.19       33.90          36.81     
  WHITE                                                
    Age                                                
      mean        37.67       31.33          33.17     
    Biomarker                                          
      afun        6.05         4.03           5.16     
B                                                      
  ASIAN                                                
    Age                                                
      mean        34.12       31.62          34.73     
  WHITE                                                
    Age                                                
      mean        39.86       39.00          34.75     
    Biomarker                                          
      afun        5.57         4.43           5.77     
C                                                      
  ASIAN                                                
    Age                                                
      mean        36.21       33.00          32.39     
  WHITE                                                
    Age                                                
      mean        39.75       44.67          36.75     
    Biomarker                                          
      afun        4.66         6.57           5.17 

Not super elegant, so still curious if there is a better way?

[Melkiades]

afaik you could avoid using pruning by having afun taking care of the different final output (it would be a different list in in_rows). I thought about using a picky splitter in the split function but it would make sense only if you have the same analysis lines but you want custom splits

[danielinteractive]

Thanks @Melkiades for your help, interesting, how would avoiding the pruning look like in above example? That is, what does the afun need to return to not show the empty lines?

[Melkiades]

Thanks @Melkiades for your help, interesting, how would avoiding the pruning look like in above example? That is, what does the afun need to return to not show the empty lines?

I hope it is what you are looking for!

library(rtables)
#> Loading required package: formatters
#> 
#> Attaching package: 'formatters'
#> The following object is masked from 'package:base':
#> 
#>     %||%
#> Loading required package: magrittr
#> 
#> Attaching package: 'rtables'
#> The following object is masked from 'package:utils':
#> 
#>     str
afun <- function(df, .var, .spl_context, ...) {
  val <- .spl_context[nrow(.spl_context), "value"]
  
  if (val == "ASIAN") {
    rcell(mean(df$AGE), label = "Age", format = "xx.xx")
  } else {
    in_rows(.list = list("mean" = mean(df$AGE), "bm" = mean(df$BMRKR1)), .formats = "xx.xx")
  }
}

lyt <- basic_table() |> 
  split_cols_by("ARM") |> 
  split_rows_by("STRATA1", page_by = TRUE) |> 
  split_rows_by("RACE", split_fun = keep_split_levels(c("ASIAN", "WHITE"))) |> 
  analyze("AGE", afun)

tab <- build_table(lyt, DM)
tab
#>            A: Drug X   B: Placebo   C: Combination
#> ——————————————————————————————————————————————————
#> A                                                 
#>   ASIAN                                           
#>     Age      32.19       33.90          36.81     
#>   WHITE                                           
#>     mean     37.67       31.33          33.17     
#>     bm       6.05         4.03           5.16     
#> B                                                 
#>   ASIAN                                           
#>     Age      34.12       31.62          34.73     
#>   WHITE                                           
#>     mean     39.86       39.00          34.75     
#>     bm       5.57         4.43           5.77     
#> C                                                 
#>   ASIAN                                           
#>     Age      36.21       33.00          32.39     
#>   WHITE                                           
#>     mean     39.75       44.67          36.75     
#>     bm       4.66         6.57           5.17

^{Created on 2025-02-03 with reprex v2.1.1}

[Melkiades]

If you also need the additional label, then I think the best way is the one you suggested

[danielinteractive]

Thanks @Melkiades , yeah in my case I have a natural split between two afuns respectively layout generating functions, therefore I cannot easily use the second way you showed. But it is not that bad with the pruning step.

Of course I wished that rtables would have a more nuanced nesting interface, such that it would not just be possible to nest or to not nest, but to say at which level to nest, which would be helpful here.

[Melkiades]

Thanks @Melkiades , yeah in my case I have a natural split between two afuns respectively layout generating functions, therefore I cannot easily use the second way you showed. But it is not that bad with the pruning step.

Of course I wished that rtables would have a more nuanced nesting interface, such that it would not just be possible to nest or to not nest, but to say at which level to nest, which would be helpful here.

I understand. In practice, it would need to have an internal "trim" (which is there btw, deprecated and not a prune) for the row split so that it would lose branches without leaves. I would not change how nested=T/F works but add a trim/prune option to the splits. Nonetheless, I think your solution is already very practical to do that

[Melkiades]

Closing this for now as it would involve a lot effort that is not planned for now. Please feel free to reopen if there is badwidth and if it is planned

[danielinteractive]

I just ran into this topic again and found my own issue again 😄
I do think there would be value here to generalize from nested = TRUE / FALSE to nested = -1 or something like that. This would save users quite some work.

[gmbecker]

This is possible and I understand the desires but it will be a lot of work. I'll have to look at my bandwidth in the new year.

…

On Tue, Nov 25, 2025, 1:00 AM Daniel Sabanes Bove ***@***.***> wrote: *danielinteractive* left a comment (insightsengineering/rtables#985) <#985 (comment)> I just ran into this topic again and found my own issue again 😄 I do think there would be value here to generalize from nested = TRUE / FALSE to nested = -1 or something like that. This would save users quite some work. — Reply to this email directly, view it on GitHub <#985 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAG53MPXOX27TGV5GKX2RCT36QLDXAVCNFSM6AAAAACNDYWBBOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTKNZUGUYTENZUHE> . You are receiving this because you were mentioned.Message ID: ***@***.***>