A note on the robustness of a full Bayesian method for nonignorable missing data analysis

Abstract

A full Bayesian method utilizing data augmentation and Gibbs sampling algorithms is presented for analyzing non-ignorable miss- ing data. The discussion focuses on a simplied selection model for re- gression analysis. Regardless of missing mechanisms, it is assumed that missingness only depends on the missing variable itself. Simulation re- sults demonstrate that the simplied selection model can recover re- gression model parameters under both correctly specied situations and many mis-specied situations. The method is also applied to analyzing a training intervention data set with missing data.