Limit tbl_uvregression() output based on p-value threshold

Question

I am running a univariate logistic regression for a model feature selection and want to be able to only display significant variables in a chart. Is there an argument within tbl_uvregression() that would allow me to only display variable that have a p-value < 0.05. Code below:

tbl <- df %>%
  tbl_uvregression(
    method = glm,
    y = diabetes,
    exponentiate = TRUE 
  )%>%
  add_global_p() %>%
  bold_p()

I tried to use the include argument to manually set the variables that were bolded (so as to signify variables with a p-value < 0.05). Surely there's a more efficient way.

Answer 1

You can modify the table in returned tbl_uvregression object, either directly through tbl$table_body or by using gtsummary::modify_table_body().
Or just use gtsummary::filter_p() :

library(gtsummary)

tbl <- mtcars[6:11] %>% 
  tbl_uvregression(
    method = glm,
    y = vs,
    method.args = list(family = binomial),
    exponentiate = TRUE
  ) %>% 
  add_global_p() %>%
  bold_p() 

tbl %>% 
  as_kable(format = "pipe")

Characteristic	N	OR	95% CI	p-value
wt	32	0.15	0.03, 0.49	<0.001
qsec	32	22.7	4.05, 582	<0.001
am	32	2.00	0.48, 8.76	0.3
gear	32	1.79	0.68, 5.14	0.2
carb	32	0.27	0.09, 0.58	<0.001

tbl %>% 
  filter_p() %>% 
  as_kable(format = "pipe")

Characteristic	N	OR	95% CI	p-value
wt	32	0.15	0.03, 0.49	<0.001
qsec	32	22.7	4.05, 582	<0.001
carb	32	0.27	0.09, 0.58	<0.001

^{Created on 2023-09-02 with reprex v2.0.2}