Identification problem of transition models for repeated measurement data with nonignorable missing values

In this paper, we consider a transition model on a response variable to describe repeated measurement data and we provide sufficient conditions to check model identifiability when analyzing data with nonignorable missing values. The sufficient conditions can give us intuitive model characteristics to achieve identifiability. In addition to the model assumptions on the response variable, a parametric model of the missing-data mechanism is often assumed. In this article, we consider identifiability in two situations: (i) both the response variable distribution and the missing-data mechanism are parametric; (ii) one of them is nonparametric, i.e., the global model is semiparametric. Useful identifiable models are proposed on the basis of these conditions. We also present an application to data of a comparative trial of two dosages of depot medroxyprogesterone acetate.