Bayesian Latent Propensity Score Approach for Average Causal Effect Estimation Allowing Covariate Measurement Error

• Activity Number: 629
• Type: Invited
• Date/Time: Thursday, August 7, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Health Policy Statistics Section
• Author(s): Yi Huang*+ and Elande Baro and Anindya Roy
• Companies: University of Maryland Baltimore County and University of Maryland Baltimore County and University of Maryland Baltimore County
• Keywords: Bayesian causal Inference ; Nondifferential Measurement Error ; Propensity Score Subclassification ; Balancing Criterion ; Finite mixture model ; Dirichlet Process

Abstract

The covariates are often measured with unobserved error in biomedical and policy studies, which is a violation of the strong ignorability assumption. The naive approach is to ignore the error and use the observed covariates in current propensity score framework for average causal effect (ACE) estimation. In the past, we showed that the naive approach typically produces biased ACE inference based on the extended causal assumptions allowing covariate errors. Based on the finite mixture model framework and the associated EM algorithm for ACE estimation for continuous outcomes in the past, we propose a new Bayesian estimation method for this joint latent propensity score approach capturing the uncertainty in propensity score subclassification arising from the unobserved measurement error as well as its influence on ACE estimation. Simulations studies are presented to show the performance of this newly developed approach vs. the existing approaches. This is a joint work with Elande Baro and Anindya Roy.