Marginalized models for longitudinal ordinal data with nonignorable dropout

Abstract

In longitudinal studies, missing data were commonly occurred. Since the missing data result in biased estimates of the estimated effects of covariates, they must be considered carefully. In this paper, we consider a selection model approach to deal with the missing data in longitudinal studies. Especially, we analyze longitudinal ordinal data with missing not at random dropouts using joint models of marginalized models and selection models. The proposed methods are illustrated using data from a longitudinal cancer clinical trial.