Pairwise likelihood inference for multilevel latent Markov models

We propose an extended latent Markov model for categorical longitudinal data with a multilevel structure. This extension allows us to take into account the correlation which may arise between the responses provided by individuals belonging to the same cluster and to model the cluster effect in a dynamic fashion. Given the complexity of computing the manifest distribution, we make inference on the model through a composite likelihood function based on all the possible pairs of subjects within every cluster. The resulting approach is illustrated through an application to a dataset concerning a sample of Italian workers.