How to manipulate the row name in tbl_regression?

Question

in the example online, there is:

mod1 <- glm(response ~ trt + age + grade, trial, family = binomial)

t1 <- tbl_regression(mod1, exponentiate = TRUE)

It produces a nice regression table that works, but how can I write the code for just showing Grade 1 instead of all of the Grades 1 through 3. And, if there was a row that was inherently binary (0 or 1), how can I choose just the true one?

I tried the label = list(.....) and value = list(...), but this is not an option I saw on the package information for gtsummary, and did not work when I tried. There must be an easy way to do this or I am not searching hard enough in the write-up. Thanks!

Answer 1

Question 1: Grade is a variable with three levels and you want to show this result on a single line. You can use the Wald or Likelihood ratio test to test for the significance of Grade altogether and print the single p-value using the combine_terms() function. Note that you will no longer see the 2 beta coefficients associated with Grade II and Grade III.

Question 2: The tbl_regression() model will print the results as they were entered into glm(). Numeric variables are interpreted as continuous and print on a single row. All others print on multiple rows. If you have a variable coded as 0/1, it will print on a single row. To show both levels, you can add factor() around it. If you are in the opposite situation where you have a binary variable printing on multiple rows and want it shown on a single row, you can use the tbl_regression(show_single_row=) argument.

Example 1: Show Grade on a single row, with default printing for "trt" (character) and "death" (numeric 0/1). "trt" prints on two rows, and "death" prints on a single row.

library(gtsummary)
glm(response ~ trt + grade + death, trial, family = binomial) %>%
  tbl_regression(exponentiate = TRUE) %>%
  combine_terms(formula_update = . ~ . - grade, test = "LRT")

Example 2: Now we will print "trt" on a single row, and "death" prints on two rows.

glm(response ~ trt + factor(death), trial, family = binomial) %>%
  tbl_regression(
    exponentiate = TRUE,
    show_single_row = trt,
    label = list(
      trt ~ "Drug B vs A (reference group)",
      `factor(death)` ~ "Death"
    )
  )