update mediation function

DESCRIPTION

@@ -41,5 +41,6 @@ Imports:
 	did (>= 2.1.2),
 	synthdid (>= 0.0.9),
 	plm (>= 2.6.3),
+	MASS,
 	foreach (>= 1.5.2)
 VignetteBuilder: knitr

NAMESPACE

@@ -9,6 +9,7 @@ export(ama_theme)
 export(balance_assessment)
 export(balance_scatter_custom)
 export(get_balanced_panel)
+export(med_ind)
 export(nice_tab)
 export(panel_estimate)
 export(plot_PanelEstimate)

R/med_ind.R

@@ -0,0 +1,54 @@
+#' Estimate Mediation Indirect Effects
+#'
+#' @description
+#' `med_ind` estimates the indirect effects of an independent variable on a dependent variable 
+#' through a mediator using Monte Carlo simulations (Selig & Preacher, 2008). It calculates the distribution of the product 
+#' of path coefficients (a*b) and provides confidence intervals for the indirect effect, along 
+#' with a ggplot histogram for visualization.
+#'
+#' @references
+#' Selig, J. P., & Preacher, K. J. (2008, June). Monte Carlo method for assessing mediation: 
+#' An interactive tool for creating confidence intervals for indirect effects [Computer software]. 
+#' Available from http://quantpsy.org/.
+#' 
+#' @param a The regression coefficient for the effect of the independent (causal) variable on the mediator.
+#' @param b The regression coefficient for the effect of the mediator on the dependent (outcome) variable.
+#' @param var_a The variance of the coefficient a.
+#' @param var_b The variance of the coefficient b.
+#' @param cov_ab The covariance between coefficients a and b.
+#' @param ci The confidence interval width for the indirect effect (default is 95 for a 95% CI).
+#' @param iterations The number of iterations for the Monte Carlo simulation (default is 20000).
+#' @param seed The seed for random number generation to ensure reproducibility (default is 1).
+#' @param theme Custom theme that follows ggplots2 (default is AMA style)
+#' @return A list containing the lower quantile, upper quantile, raw simulation data, and histogram plot of the indirect effects.
+#' @examples
+#' \dontrun{
+#' result <- med_ind(a = 0.5, b = 0.7, var_a = 0.04, var_b = 0.05, cov_ab = 0.01)
+#' result$lower_quantile
+#' result$upper_quantile
+#' result$plot
+#' }
+#' @export
+med_ind <- function(a, b, var_a, var_b, cov_ab, ci = 95, iterations = 20000, seed = 1, theme = causalverse::ama_theme()) {
+  set.seed(seed)
+
+  acov_matrix <- matrix(c(var_a, cov_ab, cov_ab, var_b), 2, 2)
+
+  sim_data <- MASS::mvrnorm(iterations, mu = c(a, b), Sigma = acov_matrix)
+  ab_values <- sim_data[, 1] * sim_data[, 2]
+
+  lower_q <- quantile(ab_values, (1 - ci/100)/2)
+  upper_q <- quantile(ab_values, 1 - (1 - ci/100)/2)
+
+  plot <- ggplot(data.frame(ab_values), aes(x = ab_values)) +
+    geom_histogram(
+      fill = "skyblue", color = "black"
+      ) +
+    labs(x = "Indirect Effect", y = "Frequency", title = "Distribution of Simulated Indirect Effects") +
+    geom_vline(xintercept = lower_q, color = "red", linetype = "dashed") +
+    geom_vline(xintercept = upper_q, color = "red", linetype = "dashed") +
+    geom_vline(xintercept = 0, color = "black", linetype = "solid") +
+    theme
+
+  list(lower_quantile = lower_q, upper_quantile = upper_q, raw_data = ab_values, plot = plot)
+}

R/synthdid_est.R

@@ -89,7 +89,7 @@ synthdid_est <-
 
     # Remove units with missing data in a time period
     if (any(is.na(data))) {
-      data <- as.data.table(data) %>%
+      data <- data.table::as.data.table(data) %>%
         .[, nas := min(get(outcome_var), na.rm = TRUE), by = !!unit_id_var] %>%
         na.omit() %>%
         .[, nas := NULL] %>%

R/synthdid_est_ate.R

@@ -115,7 +115,7 @@ synthdid_est_ate <-
 
     # Process each adoption cohort
     for (adoption_cohort in adoption_cohorts) {
-      cat("adoption_cohort:",adoption_cohort,"\n" )
+      cat("Adoption Cohort:",adoption_cohort,"\n" )
 
       # Prepare balanced panel data for current adoption cohort
       # Specifically, create balanced panel with lags and leads for all treated

docs/pkgdown.yml

@@ -5,5 +5,5 @@ articles:
   a_introduction: a_introduction.html
   b_synthdid: b_synthdid.html
   c_did: c_did.html
-last_built: 2024-01-12T21:00Z
+last_built: 2024-01-23T01:01Z