R/generate_simulated_data.R

In gapclosing: Estimate Gaps Under an Intervention

#' Generate simulated data
#' @description Generates simulated data to illustrate the gapclosing function
#' @param n Number of observations to be generated
#' @return A data frame with \code{n} rows and 4 columns containing simulated data containing \code{category} over which disparities are defined, a \code{confounder} that affects treatment assignment, a binary \code{treatment}, and a continuous \code{outcome}.
#' @references Lundberg I (2021). "The gap-closing estimand: A causal approach to study interventions that close disparities across social categories." Sociological Methods and Research. Available at {https://osf.io/gx4y3/}.
#' @export

generate_simulated_data <- function(n = 1000) {
  # Initialize non-standard evaluation variables to avoid R CMD check warnings.
  confounder <- treatment <- treatment_effect <- outcome <- NULL
  data.frame(category = sample(c("A","B","C"), n, replace = TRUE)) %>%
    # Create a confounding variable called confounder
    dplyr::mutate(confounder = stats::rnorm(n,
                                            mean = dplyr::case_when(category == "A" ~ -1.5,
                                                                    category == "B" ~ 0,
                                                                    category == "C" ~ 1.5),
                                            sd = 1),
                  # Assign treatment as a function of the confounder
                  treatment = stats::rbinom(n, size = 1, prob = stats::plogis(confounder)),
                  # Create a treatment effect that varies by category
                  treatment_effect = dplyr::case_when(category == "A" ~ 1.3,
                                                      category == "B" ~ 0,
                                                      category == "C" ~ -1.2),
                  # Generate the outcome
                  outcome = stats::rnorm(n, mean = confounder + treatment*treatment_effect, sd = 1)) %>%
    # Remove the treatment effect, which one would not know in reality
    dplyr::select(-treatment_effect)
}