Investigating the Sensitivity of Growth Mixture Model Results to Covariate-Informed Scoring

Growth mixture models (GMMs) are commonly used to identify heterogeneous developmental trajectories in longitudinal data. When GMMs are estimated using a two-step approach—first scoring instruments via item response theory (IRT) and then fitting mixture models—scoring decisions can influence class enumeration and growth parameter estimates. An unresolved methodological question is whether covariates should be incorporated during the IRT scoring stage. Using Monte Carlo simulations and an empirical application, we examined whether covariate-informed scoring improves class enumeration and parameter recovery relative to unconditional scoring. Results indicate that covariate-informed scoring provides an advantage when covariates influence the measurement model.