Description
Multilevel latent class analysis (MLCA) was developed more than a decade ago to address nested data, such as individuals nested within a higher-order structure (e.g., students nested in classrooms). By applying MLCA to the analysis of intensive longitudinal data (ILD), we can better understand the heterogeneity of behavior patterns in daily life and identify within-person vs. between-person risk factors. Specifically, we can model comprehensive, within-day substance use patterns, incorporate day-level psychosocial and person-level predictors of patterns, and predict acute (e.g., next-day) outcomes.
Modeling Latent Class Variables in the Context of Intensive Longitudinal Data (ILD)
Marginal Modeling Approach to MLCA
Two-Level Modeling Approach to MLCA
with latent classes at both levels
Marginal models (e.g.., generalized estimating equations applies to longitudinal data), are a popular alternative to random effects models for repeated-measures data. Marginal models estimate average effects in a population and produce cluster-adjusted, robust standard errors; this can be an attractive alternative to random effects models when estimating complex latent variable models.
Software resources
This approach to random effects modeling employs latent class variables at the day-level and person-level. The person-level classes summarize the random effects of interest and reflect heterogeneity across persons in their probabilities of having certain types of days. In some of the literature, this is referred to as the “non-parametric” approach to MLCA. This approach to MLCA is emerging as particularly useful in empirical data.
Software resources
Resources
Resources are being developed to help you learn more about MLCA. Here are a few introductory resources to get you started:
Advanced Models: MLCA
Let’s stay in touch.
We are in this together. Receive an email whenever a new model or resource is added to the Knowledge Base.