| This study empirically investigated the issues in growth mixture modeling concerning the capacity of undercovering the correct number of latent groups in a heterogeneous population and parameter estimates. A Monte Carlo simulation design was used to systematically manipulate the three design factors: heterogeneity levels, multivariate distance, and sample size.;The results of this study showed that the growth mixture model is not effective in detecting the correct number of latent groups even at large multivariate distances between groups, and tends to over-extract the number of groups in a population. This is especially apparent when the sample size is large. Additionally, the model is not effective to recover the original proportions of groups in the population and resulted considerable bias in fixed effect and variance parameter estimates. BIC, AIC, and entropy were found to be unreliable as indices for model selection and group classification. |
