You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Default rounding in R is different from SAS and from what everyone expects.
Considered Options
Export function to implement standard rounding
Rely on default R rounding, and continue to explain to people that the formatted values are "correct" in the context of base R's default rounding
Decision Outcome
After discussion with @shajoezhu, we decided we will export a function for human rounding, so we don't have to continue to explain how the rounding is performed and results from other languages will match.
In some cases, tabulations and summaries need to be calculated by every combination of variables, including observed and unobserved combinations (e.g. tabulating subject characteristics by treatment arm: even if a characteristic is unobserved in a treatment arm), and in other cases, we need summaries by observed levels (e.g. tabulating preferred term stratified by SOC).
Considered Options
We already export functions for summarizing by all combinations (ard_categorical(by)) and we could add a strata argument to tabulate by observed levels.
Instead of adding an argument to the existing function, we could export another function that can do stratified summaries.
Decision Outcome
@bzkrouse and @ddsjoberg discussed and we will add a strata argument to supplement the existing by argument.
https://github.com/insightsengineering/chevron.adr
For now, each comment below will be one decision made.
The text was updated successfully, but these errors were encountered: