Feature request: Association Tests Between Two Categorical Variables (Both With Multiple Response Options) about gtsummary HOT 4 CLOSED

Thanks for the post @rawand-hanna . I think you'll need to add some concrete examples of what you're requesting. I am not sure I follow 100%

from gtsummary.

Hey @ddsjoberg
Thanks for the response

Here is an example

library(tidyverse)
library(gtsummary)

df <- tibble(language = c('Arabic French', 'French Spanish', 'Spanish Arabic', 'Arabic', 'French'),
education = c('Primary', 'Primary', 'Secondary', 'Secondary', 'Primary'))

df %>% tbl_summary(by = education) %>%
add_overall() %>%
add_p()

Here is the result

The issue is that the cross table is not displaying each language on a separate row, instead, it displays each combination of languages on each row.

Is there a way to tell gtsummary that the type of a variable is categorical with 'select multiple' type?

from gtsummary.

Perhaps this will be helpful: https://mskcc-epi-bio.github.io/bstfun/reference/add_variable_grouping.html

As I mentioned in the stackoverflow post you recently made, gtsummary will only tabulate what is in the data frame.

from gtsummary.

Thanks @ddsjoberg
You are the best

Just a quick note, while using the add_variable_grouping() function, the p-value is still calculated per each of the categorical variables with the cross-tab variable.

I used the same sample code as here

Added age_group variable to the code

`set.seed(11234)

data.frame(
race_asian = sample(c(TRUE, FALSE), 20, replace = TRUE),
race_black = sample(c(TRUE, FALSE), 20, replace = TRUE),
race_white = sample(c(TRUE, FALSE), 20, replace = TRUE),
age = rnorm(20, mean = 50, sd = 10)
) %>%
mutate(age_group = cut(age, breaks = 3)) %>%
gtsummary::tbl_summary(by = age_group,
label = list(race_asian = "Asian",
race_black = "Black",
race_white = "White",
age = "Age")
) %>%
add_variable_grouping(
"Race (check all that apply)" = c("race_asian", "race_black", "race_white")
) %>%
add_p()`

from gtsummary.