feat: add test_catcat

R/test_catcat.R

@@ -0,0 +1,79 @@
+#' @title 
+#' Compute chi-square or Fisher's exact test for two categorical variables
+#'
+#' @description 
+#' This function computes a chi-square or Fisher's exact test for two categorical variables in a data frame.
+#'
+#' @param data A data frame containing the variables of interest.
+#' @param x A character string specifying the name of the first variable.
+#' @param y A character string specifying the name of the second variable.
+#' @param na_x A vector of values to be treated as missing in \code{x}.
+#' @param na_y A vector of values to be treated as missing in \code{y}.
+#' 
+#' @details 
+#' If the cell counts are lower than 5, the function will use Fisher's exact test. Otherwise, it will use a chi-square test.
+#'
+#' @return A tibble containing the results of the chi-square or Fisher's exact test.
+#'
+#' @examples
+#' data("mtcars")
+#' test_catcat(mtcars, "cyl", "vs")
+#'
+#' @importFrom rstatix chisq_test
+#' @importFrom stats fisher.test
+#' @importFrom dplyr filter mutate select
+#' @importFrom tidyr pivot_longer
+#' @importFrom broom tidy
+#'
+#' @export
+test_chisq <- function(data, x, y, na_x = NULL, na_y = NULL){
+
+ # remove NA values
+ data2 <- 
+ data %>%
+ filter(!(!!sym(x) %in% na_x)) %>%
+ filter(!(!!sym(y) %in% na_y))
+
+ # Create new variables to feed into `rstatix::chisq_test()`
+ stat_x <- data2[[x]]
+ stat_y <- data2[[y]]
+
+ # Check expected cell counts
+ expected_counts <-
+ chisq.test(table(data2[[x]], data2[[y]]))$expected %>%
+ suppressWarnings()
+
+ if (any(expected_counts < 5)) {
+ # Use Fisher's exact test if expected cell counts are low
+ result <- fisher.test(x = factor(stat_x), y = factor(stat_y)) %>%
+ broom::tidy(out) %>% # Return a data frame
+ mutate(n = NA,
+ statistic = NA,
+ df = NA,
+ `p.signif` = NA,
+ p = `p.value`) %>%
+ select(
+ n, 
+ statistic,
+ p, 
+ df,
+ method, 
+ p.signif,
+ alternative
+ )
+
+ } else {
+ # Use chi-square test if expected cell counts are not low
+ result <- rstatix::chisq_test(x = stat_x, y = stat_y) %>%
+ mutate(alternative = NULL)
+ }
+
+ # Return results
+ dplyr::tibble(
+ col_x = x,
+ col_y = y
+ ) %>%
+ cbind(result) %>%
+ dplyr::as_tibble()
+
+}