Two Monte Carlo studies for latent class segmentation models

Abstract

Model assessment and comparison are essential aspects of statistical inference. The likelihood ratio test is one of the main instruments for model selection; however, this is not appropriate when the model under consideration contains random effects. In this paper, we present two simulation studies for latent class segmentation models. The first Monte Carlo study compares the performance of seven Information Criteria in predicting the correct number of segments. The second study investigates factors that have an effect on segment membership and parameter recovery and affect computational effort.