We illustrate a class of conditional models for the analysis of longitudinaldata suffering attrition in random effects models framework, where thesubject-specific random effects are assumed to be discrete and to follow atime-dependent latent process. The latent process accounts for unobservedheterogeneity and correlation between individuals in a dynamic fashion, and fordependence between the observed process and the missing data mechanism. Ofparticular interest is the case where the missing mechanism is non-ignorable.To deal with the topic we introduce a conditional to dropout model. A shapechange in the random effects distribution is considered by directly modelingthe effect of the missing data process on the evolution of the latentstructure. To estimate the resulting model, we rely on the conditional maximumlikelihood approach and for this aim we outline an EM algorithm. The proposalis illustrated via simulations and then applied on a dataset concerning skincancers. Comparisons with other well-established methods are provided as well.
展开▼