A Monte Carlo Simulation Study on Model Selection in Latent Markov Models

Latent Markov models emerge as a good alternative for longitudinal psychological studies where it is not possible to take quantitative measurements to analyze and interpret time-dependent changes of categorical observed and latent variable(s). However, despite its increasing use in recent years, a consensus on the model selection process in the Latent Markov models has not been reached yet In this context, the first objective of this research was to provide an application example by using an empirical dataset with a single variable. Another aim was to examine the impacts of the strength of item response probabilities, the number of times the measurement being taken and sample size on model selection and parameter estimation bias based on using the dataset generated by Monte Carlo simulation method. As a result, using BIC and CAIC information criteria, 100% correct decision rate was observed regardless to item response probabilities (weak or strong), and number of measurement (2 or 3) when the sample size increased from 200 to 2000. The findings were discussed in the light of the related literature.