as_result_df diffs noticed in ARD output

[ayogasekaram]

In creating the documentation for QCing rtables with ARDs, there are some differences noticed that can be address if improving the output.

• 1. group1 and group1_level don't match the same columns in the ARD
• 2. There are 2 rows labeled Mean(SD), this could be separated to match the actual value.
• 3. The AGE variable has an extra "A" in the front (the data does not have this extra A)
• 4. "context" column is missing
• 5. Some items in the variable_level column have the stat names.

[Melkiades]

1. group1 and group1_level don't match the same columns in the ARD

This is something present in {rtables}. I added an option (turned off by default) to have the table_name split or not

3. There are 2 rows labeled Mean(SD), this could be separated to match the actual value.

variable_label is a column that is not present in {cards} and only shows the final label. There are 2 because one row is mean, the other is sd if you check stat values. This can be stat_label, if it makes sense

5. The AGE variable has an extra "A" in the front (the data does not have this extra A)

I cannot reproduce this bug so far. Could you copy the used code?

7. "context" column is missing

In my extracted ARD from {gtsummary} this is missing. Do we need the context column? @shajoezhu @khatril @ddsjoberg. afaik it would complex to extract only for special cases such regression

9. Some items in the variable_level column have the stat names.

I think this is something we need to discuss. At the moment, as_result_df pulls the row name as variable name because {rtables} uses often statistics as final variables. Probably, we need to use the last split value for these cases. I am open to propositions

[ddsjoberg]

Agreed, I don't think we need the "context" column in the flattened rtable. The stat names should be enough for comparing results.

[ayogasekaram]

Hey @Melkiades

I pulled the recent rtables changes and ran the code but am getting this error

Here is the code I'm using for the QCing Document

Create a table using the chevron package

library(chevron)
dmt01 <- chevron::run(dmt01, syn_data)
results <- rtables::as_result_df(dmt01, make_ard = TRUE, expand_colnames = TRUE)

Create a comparable ARD

library(cards)
ards <- ard_stack(syn_data$adsl, ard_continuous(variables = c(AGE), statistic = ~ continuous_summary_fns(c("N", "mean", "sd", "median", "min", "max"))
  ),
                 ard_categorical(variables = c(AGEGR1, SEX, ETHNIC, RACE)),
                 .by = "ARM",
                 .overall = TRUE)

[Melkiades]

Update on AAGE. This is coming from {chevron}. It is enough to change ARD to reflect that (check chevron::dmt01_main for the default variables). @BFalquet fyi

> chevron::syn_data$adsl$AAGE
 [1] 32 35 30 26 40 49 34 32 24 24 40 24 40 28 37 40 40 38 34 45 44 34 38 42 26 31 39 36
[29] 28 30 34 34 26 27 32 27 57 40 31 30 35 39 25 43 25
attr(,"label")
[1] "Age (yr)"
> chevron::syn_data$adsl$AGE
 [1] 32 35 30 26 40 49 34 32 24 24 40 24 40 28 37 40 40 38 34 45 44 34 38 42 26 31 39 36
[29] 28 30 34 34 26 27 32 27 57 40 31 30 35 39 25 43 25
attr(,"label")
[1] "Age"

[shajoezhu]

I don't think we need the context column for now. However, I think if want to reconstruct the rtables from dataframes, we would want to use the context column? I m not sure. anyway, let's remove it for now.