Bayesian Jackknife Empirical Likelihood Based Inference for Missing Data and Causal Inference Problems

• Activity Number: 25 - Modern Techniques in Handling Missing Data with Challenging Data Structures Including Big and Small Data Files
• Type: Topic Contributed
• Date/Time: Monday, August 3, 2020 : 10:00 AM to 11:50 AM
• Sponsor: Survey Research Methods Section
• Author(s): Yichuan Zhao* and Sixia Chen and Yuke Wang
• Companies: Georgia State University and University of Oklahoma Health Sciences Center and Georgia State University
• Keywords: Empirical Likelihood; Bayesian Inference; Jackknife; Missing Data; Causal Inference

Abstract

Missing data is very common in variety of fields including survey sampling, economics, social science, and medical researches. Missing data reduces the representativeness of the sample and could potentially lead to inference problems. Empirical likelihood (EL) method, considered by Owen (2001), does not require any distributional assumptions. It is a powerful tool for statistical inference for data with missing values. This study applied the Bayesian jackknife empirical likelihood method, proposed by Cheng and Zhao (2019), for inference with missing data and causal inference.

The propensity score weighted estimator, doubly robust estimator, and semiparametric fractional imputation estimator, proposed by Chen and Kim (2017), were used for inference with missing data. Some existing methods were compared with the Bayesian jackknife empirical likelihood approach in a simulation study and the proposed approach shows better performance in many scenarios. And a casual inference example with real data was used to illustrate the proposed approach.