Three-step estimation of latent Markov models with covariates

We propose a modified version of the three-step estimation method for the latent class model with covariates, which may be used to estimate a latent Markov (LM) model with individual covariates and possible dropout. We illustrate the proposed approach through an application finalized to the study of the health status of elderly people hosted in Italian nursing homes. This application is based on an LM model which accounts for informative missing responses and dropout due to the death of a patient. In the analysis of this dataset, which is also characterized by a large number of response variables and covariates, we found the proposed three-step approach more stable with respect to the standard maximum likelihood method.