Propensity Score Analysis with Unreliable Covariates: A comparison of Five Reliability-adjustment Methods

• DOI: 10.31234/osf.io/yk2b7
• Publication Date: 2023-07-18T05:01:04Z
• Authors: Huibin Zhang; Walter Leite
• Keywords: Quantitative Methods; Statistical Methods; Social and Behavioral Sciences

Propensity score analysis (PSA) is often used by researchers to control for selection bias due multiple covariates in quasi-experimental studies. However, with low reliability have been shown lead biased treatment effects estimates PSA. Latent variable a promising strategy reduce the negative of observed variables’ measurement error. This Monte Carlo simulation study compared performance five methods adjust propensity scores unreliability. The results indicate that latent model inclusive factor (PSIF) generated lowest relative effect estimates, followed estimation structural equation (PS-SEM). only PSIF provided unbiased across conditions high, medium and reliability. also show evaluation covariate balance can be misleading when there are unreliable covariates, because balanced deemed adequate.