loop over results from logistic regression and put them in a data frame-CodePudding

I have an issue with looping over results of three logistic regression analyses. I am trying to put the exp(b)/odds ratio and their 95% confidence intervals into a data frame, side-by-side. I would like to have the model name (e.g., model 1, model 2, and model 3) in the columns and the variables on the rows. I have been trying to modify code I have found from this site--(I thought I would check here before I would post my question, but nothing is spot on, but if I missed it please point me in that direction). Unfortunately, this appears to be a bit above my knowledge base. This is the code that I am trying to use. It does the base calculations, and creates a tibble with just the coefficients, but I need the exp(b)/odds ratio and the CIs for each model as I said with the model name in the columns, and the variables in the rows.

library(tidyverse)
mydata <- read.csv("https://stats.idre.ucla.edu/stat/data/binary.csv")

forms <- list(model1 = admit ~ gre, model2 = admit ~ gre   gpa, model3 = admit ~ gre   gpa   rank)

map_df(forms, ~coef(glm(.x, mydata, family = "binomial"))) %>%
  select(-1)

How would I modify the script to collect the information I need, and in the data frame style my study will need?

CodePudding user response：

broom::tidy() is very handy for these problems. Try

m1 <- (forms
   |> map(glm, data = mydata, family = "binomial")
   |> map_dfr(broom::tidy, exponentiate = TRUE, conf.int = TRUE, .id = "model")
)
print(m1)

select() and pivot_wider() are useful for converting, e.g.

(m1 
   |> dplyr::select(model, term, estimate, conf.low, conf.high) 
   |> pivot_wider(names_from = "term",
                  values_from = c("estimate", "conf.low", "conf.high"))
)