Latent Class Modeling in the Context of Intensive Longitudinal Data: Estimation Considerations

Google map

Abstract: 

Intensive longitudinal data are characterized by repeated assessments of numerous facets (e.g., mood, behavior, context). Multilevel latent class analysis (MLCA) expands the capabilities of traditional multilevel modeling to examine patterns at each moment/day and incorporate person- and moment/day-level predictors of latent class membership. We discuss competing approaches for MLCA estimation, including marginal models with robust standard errors and random effects using a fully parameterized model, a common factor model, and a Level-2 latent class variable. Software considerations are discussed and a web-based app to facilitate interpretation of model parameters is demonstrated.