THREE-STEP RECTANGULAR LATENT MARKOV MODELING BASED ON ML CORRECTION

Rectangular latent Markov (LM) models have been recently introduced to account for different numbers of latent states over time. This contribution proposes a three-step estimation procedure for such models, which proved useful in the LM modeling framework for flexibility. Specifically, a bias-adjusted maximum likelihood (ML) estimator is introduced for the third step. A simulation study provided preliminary encouraging results regarding the efficacy and effectiveness of the method.