Modeling of Continuous-Count and Continuous-Ordinal Mixed Longitudinal Bivariate Inflated Responses with Possibility of Non Random Missingness

Abstract

One of the most frequently encountered longitudinal studies issues is data with losing the appointments or getting censoring. In such cases, all of the subjects do not have the same set of observation times. The missingness in the analysis of longitudinal discrete and continuous mixed data is also common, and missing may occur in one or both responses. Failure to pay attention to the cause of the missing (the mechanism of the missingness) leads to unbiased estimates and inferences. Therefore, in this paper, we investigate the mechanism of nonignorable missing in set-inflated continuous and zero-inflation power series, as well as the continuous and k -inflated ordinal mixed responses. A full likelihood-based approach is used to obtain the maximum likelihood estimates of the parameters of the models. In order to assess the performance of the models, some simulation studies are performed. Two applications of our models are illustrated for the American's Changing Lives (ACL) survey, and the Peabody Individual Achievement Test (PIAT) data set.