This repository contains the manuscript and code for paper "Simulation-Induced Bias in Quantities of Interest."
Following King, Tomz, and Wittenberg (2000) researchers commonly obtain point estimates of quantities of interest by simulating model coefficients, transforming these simulated coefficients into simulated quantities of interest, and then taking the average of the simulated quantities of interest. In contract, other researchers directly transform coefficient estimates into quantities of interest, using the invariance property of maximum likelihood estimators. These approaches are not equivalent. We show both analytically and empirically that computing quantities of interest using the average of simulations can introduce substantial bias. Instead, researchers should use the invariance property to calculate the maximum likelihood estimates of their quantities of interest.