A post-stratification method to improve the accuracy of the naive propensity score adjusted estimator in MNAR nonresponse

Nonresponse occurs frequently in sample surveys, and the appropriate method of handling nonresponse depends on the nonresponse mechanism.Many methods have been developed and used to handle MCAR (missing completely at random) or MAR (missing at random) nonresponse.However, appropriate methods for handling MNAR (missing not at random) nonresponse are still being studied.In particular, research is actively being conducted on the calibration method, which adjusts weights by using auxiliary variable information.Chung and Shin ( 2017) proposed a post-stratification method using substrata, which is one of the calibration methods.In this study, we propose a new post-stratification method to improve the accuracy of the propensity score adjusted estimator, and the superiority of the proposed method over existing methods is confirmed through simulation studies.