Speaker: Yanghyeon Cho, PhD Candidate, Department of Statistics, Iowa State University

Title: Handling Non-monotone Missing Data under Pattern Graph Model

Abstract: Non-monotone missingness is frequently encountered in real-world problems, with randomly reentering subjects who dropped out of the longitudinal study as a classic example. Previous attempts have been made to manage non-monotone missing patterns under the Missing At Random (MAR) condition. However, the MAR assumption may not adequately address the complexities of non-monotone missingness situations. To this end, we adopt the pattern graph model to identify the nonignorable missing mechanism. With the mechanism identified, this paper presents two approaches for estimating the parameter of interest. One is a fully parametric approach and the other is a model-assisted semiparametric approach.  We provide an in-depth illustration underpinned by concrete examples to elucidate these estimation procedures. Extensive simulation studies have been conducted to empirically validate the proposed methods.