Bug Report: add_stat_label() can no longer be forced to not update dichotomous or categorical labels

[dereksonderegger]

The function add_stat_label() unfortunately affects all statistics and not just the ones that the user wants to update. This was particularly annoying for dichotomous and categorical variables because changing the labels in a continuous2 type would then force a label change to add the n (%) to everything else.

In previous versions of gtsummary, we were able to work around this by setting the new label to NA_character_ and the output would look like nothing had changed for the dichotomous and categorical variables. Unfortunately this trick no longer works in version 2.0.1.

tmp <- data.frame(
  A = c(TRUE, TRUE, TRUE, FALSE),
  B = c(TRUE, FALSE, FALSE, FALSE),
  C = rnorm(4),
  D = c('G1','G2','G1','G2')
)

# Default behavior for logical or categorical is to suppress the n (%)
# in favor of the footer
tmp |>
  gtsummary::tbl_summary(
    type = list(C ~ 'continuous2'),
    statistic = list(
      gtsummary::all_continuous2() ~ c('{N_obs}','{mean} ({sd})')
    )
  )

# When we add the custom statistics label for continuous2 variables
# the n (%) now shows up on in the variable labels
tmp |>
  gtsummary::tbl_summary(
    type = list(C ~ 'continuous2'),
    statistic = list(
      gtsummary::all_continuous2() ~ c('{N_obs}','{mean} ({sd})')
    )
  ) |>
  gtsummary::add_stat_label(
    list(
      gtsummary::all_continuous2() ~ c('Number of Subjects', 'Mean (Std Dev)')
    ),
    location='row'
  )


# In previous versions of gtsummary, the solution was to set
# the label for dichotomous or categorical variables to
#  `NA_character_`, but that no longer works
tmp |>
  gtsummary::tbl_summary(
    type = list(C ~ 'continuous2'),
    statistic = list(
      gtsummary::all_continuous2() ~ c('{N_obs}','{mean} ({sd})')
    )
  ) |>
  gtsummary::add_stat_label(
    list(
      gtsummary::all_continuous2() ~ c('Number of Subjects', 'Mean (Std Dev)'),
      gtsummary::all_dichotomous() ~ NA_character_
    ),
    location = 'row'
  )

# Passing an empty string sort-of works, but leaves an obnoxious
# comma 
tmp |>
  gtsummary::tbl_summary(
    type = list(C ~ 'continuous2'),
    statistic = list(
      gtsummary::all_continuous2() ~ c('{N_obs}','{mean} ({sd})')
    )
  ) |>
  gtsummary::add_stat_label(
    list(
      gtsummary::all_continuous2() ~ c('Number of Subjects', 'Mean (Std Dev)'),
      gtsummary::all_dichotomous() ~ ''
    ),
    location = 'row'
  )

Ideally, the merge would be acting more like gt::cols_merge() where the pattern cols_merge_pattern = "{label}, {stat_label}" could be replaced with cols_merge_pattern = "{label}<<, {stat_label}>>" and updated labels that are NA will work.

[ddsjoberg]

Thanks @dereksonderegger !

Issue should be solved in the dev version of the pkg. Can you confirm, please?

[dereksonderegger]

Hi @ddsjoberg !

Thank you so much for responding to this. I've checked that the reprex example works and I've been successful using gtsummary version 2.0.1.9008 in my code stack.