Easier identification of dichotomous variables and all_categorical(), all_continuous(), all_dichotomous() helpers about broom.helpers HOT 3 CLOSED

I love this. For now, these functions have only worked with tbl_summary(), and I have wanted to generalize them for tbl_regression(). We should implement something like this! Just need to decide on the best route!

The possible types would be continuous, categorical, dichotomous, interaction, and unknown (for when there is no model.frame() method)? We should probably implement selectors for all types (well, except known)?

• all_continuous()
• all_categorical(dichotomous = TRUE)
• all_dichotomous()
• all_interaction()

I can implement the same in gtsummary, but I am not sure I'll be able to merely reexport the broom.helper functions if I want them to work with all the gtsummary obejcts. For example, we also have a summary type called continuous2 that has it's own selector. There is likely a pretty straightforward way around the issue...just need to sit with it longer and think :)

from broom.helpers.

Sure, we should think properly how to implement it.

If I understood properly, your trick in broom::.tidy_tidyselect(), you recreate a tibble (df_empty) with the list of variables to be able to apply dplyr::select().

I guess you use something similar in gt_summary.

Maybe a possible trick could be to add a "var_type" attribute to each column of df_empty. Then, to create functions as is_continous(), is_categorical() who test the value of the attribute, and finaly to code all_categorical() as a where(is_categorical)?

Just brainstorming...

Just some clarifications: what is exactly the type continuous2?

Regards

from broom.helpers.

continuous2 variables are those that are summarized on 2 or more rows in tbl_summary() and tbl_svysummary().

trial %>%
  select(age, trt) %>%
  tbl_summary(
    by = trt,
    type = all_continuous() ~ "continuous2",
    statistic = all_continuous() ~ c("{N_nonmiss}",
                                     "{mean} ({sd})",
                                     "{median} ({p25}, {p75})"),
    missing = "no"
  )

from broom.helpers.