R/cy_GenderStats.R

In vanichols/CyChecks: Obtains and Analyzes Iowa State University Salary Data

# Function to be called inside real function
fun1 <- function(data) {

  # -------Start fun
  # just to calm down R CMD CHECK
  position <- NULL
  gender <- NULL
  n <- NULL
  M <- NULL
  term <- NULL
  p.value <- NULL
  p_val <- NULL

  # Handle bad input
  assertthat::assert_that(is.data.frame(data))
  assertthat::assert_that("position" %in% names(data))
  assertthat::assert_that("gender" %in% names(data))
  assertthat::assert_that("total_salary_paid" %in% names(data))
  assertthat::not_empty(data)
  assertthat::assert_that(is.numeric(data$total_salary_paid))

  # Make sure there are two genders in that department (eye roll)
  twogenders <- data %>%
    dplyr::group_by(position, gender) %>%
    dplyr::summarise(n = n()) %>%
    dplyr::mutate(gender = as.character(gender)) %>%
    dplyr::filter(gender %in% c("F", "M")) %>%
    tidyr::spread(gender, n)

  # If there are only M/F (2 cols), assign p_val of NA
  if (ncol(twogenders) < 3) {
    myres <- tibble::tibble(term = "gender",
                    p_val = NA,
                    verdict = NA)

  } else {
    # Figure out which positions have both a M and F
    poslist <-  twogenders %>%
      dplyr::filter(!is.na(M),!is.na(`F`)) %>%
      dplyr::filter((M > 1 & `F` > 1)) %>% #--you need more than 1 of each to do a comparison
      dplyr::pull(position)

    # Filter to get only those positions
    mydsub <- data %>%
      dplyr::filter(position %in% poslist)

    # If it's empty, it needs to make a fake tibble
    # If it's not empty and has more than 1 position, fit a simple model

    #--If#1 (not empty)
    if (nrow(mydsub) > 0) {
      #--If#2 (more than 1 position)
      if (length(poslist) > 1) {
        # This is what I want it to do if mydsub isn't empty
        myres <- broom::tidy(stats::anova(
          stats::lm(total_salary_paid ~
                      position + gender, data = mydsub)
        )) %>%
          dplyr::filter(term  == "gender") %>%
          dplyr::select(term, p.value) %>%
          dplyr::rename("p_val" = p.value) %>%
          dplyr::mutate(verdict = ifelse(p_val < 0.2, "boo", "ok"))
      } else {
        # Not empty, only one position
        myres <-
          broom::tidy(stats::anova(stats::lm(total_salary_paid ~
                                               gender, data = mydsub))) %>%
          dplyr::filter(term  == "gender") %>%
          dplyr::select(term, p.value) %>%
          dplyr::rename("p_val" = p.value) %>%
          dplyr::mutate(verdict = ifelse(p_val < 0.2, "boo", "ok"))
      }
    } else {
      # It's empty
      myres <- tibble::tibble(term = "gender",
                      p_val = NA,
                      verdict = NA)
    }
  }

  # If it made it through all that and still has a funky output, overwrite it
  if (ncol(myres) < 3) {
    myres <- tibble::tibble(term = "gender",
                    p_val = NA,
                    verdict = NA)

  }

  return(myres)
}

#' Function to compare M vs F salaries within a department
#'
#' @name cy_GenderStats
#' @importFrom tidyr nest unnest
#' @importFrom dplyr group_by summarise select rename mutate filter arrange
#' @importFrom assertthat assert_that not_empty
#' @param data A dataframe of ISU salary data with academic department info. Default is for year 2018. Column names must include 'department', 'position', 'gender', and 'total_salary_paid'.
#' @return A dataframe of department, nested data, p-value for gender pay gap after accounting for position, and a verdict
#' @examples
#' data(sals18)
#' sals18 %>%
#' dplyr::filter(department == "AGRONOMY") %>%
#' dplyr::filter(stringr::str_detect(position, "PROF")) %>%
#' cy_GenderStats()
#'
#' @export

# Actual p
cy_GenderStats <- function(data = sals18){

  # Make sure it has the columns I want, and that it's not empty
  assertthat::assert_that(is.data.frame(data))
  assertthat::assert_that("department" %in% names(data))
  assertthat::assert_that("position" %in% names(data))
  assertthat::assert_that("gender" %in% names(data))
  assertthat::assert_that("total_salary_paid" %in% names(data))
  assertthat::not_empty(data)
  assertthat::assert_that(is.numeric(data$total_salary_paid))


  yourstats <- data %>%
    dplyr::group_by(department) %>%
    tidyr::nest() %>%
    dplyr::mutate(
      mod = data %>%
        purrr::map(purrr::possibly(fun1,
                                   otherwise = NA_real_))) %>%
    tidyr::unnest(mod) %>%
    dplyr::arrange(p_val)

  return(yourstats)
}