Sensitivity and Robustness of Propensity Score Estimation to the Impact of Unobserved Covariates

• In Event: Bayesian Estimation and Propensity Score Analysis in Educational Statistics

Abstract

As a crucial component of propensity score methods for reducing selection bias, propensity score estimation can only account for observed covariates. The present study introduces a new technique to assess the sensitivity and robustness of propensity score estimation to the impact of unobserved covariates. The sensitivity and robustness are assessed through a reference distribution that can be approximated to a Pearson distribution by borrowing information from observed covariates. The procedure of the assessment is illustrated with empirical data on substance abuse prevention for high-risk youth.

Authors

• Wei Pan, Duke University

• Haiyan Bai, University of Central Florida