update tidy_add_coefficients_type() for binomial model with log link

[larmarange]

@ddsjoberg

Following ddsjoberg/gtsummary#788 we could consider to update tidy_add_coefficients_type() to take into account binomial models with log link.

Such models produce Relative Risks (RR) or Prevalence Ratios (PR) estimates.

A possibility would be to have a new coefficient type "prevalence_ratio" (with the associated coefficient_label "PR" or "log(PR)"?

Or would you prefer "relative_risk" / "RR" or "log(RR)"

Would it be problematic on gtsummary side? maybe just a litte update to deal with acronym def?

[ddsjoberg]

Adding the acronym would be great. I prefer relative risk as the default I'd you're ok with that!

[larmarange]

@ddsjoberg It works OK with gtsummary. Just no acronym definition in footnote (that part is managed by gtsummary directly)

library(broom.helpers)
library(gtsummary)
#> 
#> Attachement du package : 'gtsummary'
#> The following objects are masked from 'package:broom.helpers':
#> 
#>     all_continuous, all_contrasts
mod <- glm(response ~ age + grade * trt, gtsummary::trial, family = binomial(log))
mod %>% tbl_regression(exponentiate = TRUE) 
#> Warning: The `.dots` argument of `group_by()` is deprecated as of dplyr 1.0.0.

^{Created on 2021-02-18 by the reprex package (v1.0.0)}