Average Causal Effect Estimation Allowing Covariate Measurement Error

Yi Huang, Department of Mathematics and Statistics
Xiaoyu Dong, Department of Mathematics and Statistics
Andrew Raim, Department of Mathematics and Statistics
Elande Baro, Department of Mathematics and Statistics
Dr. Karen Bandeen-Roche, Johns Hopkins Bloomberg School of Public Health
Dr. Cunlin Wang, CDER, FDA


Covariates are often measured with 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. However, after extending the existing causal framework incorporating assumptions allowing errors-in-covariates, the naive approach typically produces biased ACE inference. In this project, we develop a finite mixture model framework for ACE estimation with continuous outcomes, which captures the uncertainty in propensity score subclassification from unobserved measurement error using the joint
likelihood. The proposed approach will estimate the propensity score subgroup membership and subgroup-specific treatment effect jointly. This will extend the current propensity score subclassification approach to accommodate cases where covariates are measured with errors.