0

I ran multiple logistic regressions based on a race_category variable using dplyr's group_by, and would like to obtain odds ratios rather than the standard output (e.g., coefficients).

I have a data frame upon which I want to predict an outcome given several variables, and wanted to determine whether there are any differences by race. I used dplyr to group by race category to run a logistic regression as follows:

r_models <-  dat_hms22 %>% 
  group_by(race_category) %>% 
    do(model = glm(talk1_1 ~ stig_pcv_con + txfrf + stig_per_con + ther_help_me
                 + meds_help_me + knowwher_temp + camp_supp + percneed + ther_ever,
                 data = ., family = binomial(logit)))
r_models$model

However, I would like to report the output as effect sizes using odds ratios.

Unfortunately, running the following line of code: r_models %>% do(exp(cbind(OR = coef(.$model), confint(.$model))))

Resulted in the equivalent of a segfault, causing R Studio to reload, so any help would be appreciated.

yotiscordero
  • 1
  • 1
  • Perhaps look at [this post](https://stackoverflow.com/questions/1169539/linear-regression-and-group-by-in-r) where you can do models by group, then use broom package with `tidy`, and include exponentiate and conf.int arguments to get the desired results. – Ben Jun 28 '23 at 16:55

0 Answers0