R/medciAsymp.R

In RMediation: Mediation Analysis Confidence Intervals

# Added 3/28/14 for asymptotic method
medciAsymp <-
  function(mu.x, mu.y, se.x, se.y, rho = 0, alpha = .05, ...) {
    p <- alpha / 2
    quantMean <- mu.x * mu.y + se.x * se.y * rho
    quantSE <- sqrt(se.y^2 * mu.x^2 + se.x^2 * mu.y^2 + 2 * mu.x * mu.y * rho * se.x * se.y + se.x^2 * se.y^2 + se.x^2 * se.y^2 * rho^2)
    CI <- quantMean + c(qnorm(alpha / 2), qnorm(1 - alpha / 2)) * quantSE
    res <- list(CI, quantMean, quantSE)
    names(res) <- c(paste((1 - alpha) * 100, "% ", "CI", sep = ""), "Estimate", "SE")
    return(res)
  }