Evaluation of the Use of Propensity Scores in Mediation Analysis

• In Event: Mediation, Ordinal, and Nonlinear Modeling

Abstract

The use of propensity scores to estimate causal effects in mediation analysis is a new concept. Classical + Propensity Model (C+PM) is one of the approaches proposing the idea of using propensity scores to provide unbiased estimates of mediated effects (Coffman, 2011). This Monte Carlo simulation study evaluates the performance of C+PM by comparing it with structural equation modeling approach as a baseline method which does not include any confounder of mediator or outcome, and an adjusted regression model including confounders in the model as separate variables in mediation analysis in terms of bias. Our study aims to compare C+PM model incorporating propensity scores into the model as a covariate with a revised C+PM model using propensity score weighting algorithm.

Authors

• Yasemin Kaya, University of Florida

• Walter L. Leite, University of Florida