The Final Class Model Depends on the Index: Exploring Bayesian Model Fit Index Performance in Growth Mixture Modeling

Investigating Bayesian model selection indices in Latent Growth Mixture Modeling (LGMM), our study evaluates DIC, WAIC, and LOOIC’s performance for optimal latent class identification. Through a simulation study (varying class proportions, class separation, and sample size), we found substantively important differences in index effectiveness, informing much-needed model selection guidelines.