Gauging Growth Trajectories of Two Related Outcomes Simultaneously Using Latent Growth Modeling

This study demonstrates how to gauge growth trajectories of two related outcomes simultaneously using the parallel latent growth models (LGM). A 30-month period longitudinal study data on crack-cocaine use and BDI-defined depression scores were used for the modeling. We first evaluated the form and determinants of the growth trajectories and the association between the trajectories of the two outcomes. Then, we scrutinized the changes in each outcome in specific time periods of interest during the entire observation period. Differences in outcome change in specific sub-time periods between ethnic groups were tested. The standard errors of the model predicted mean outcome changes in specific time periods were estimated via both the delta method and bootstrapping. Mplus and its RUNALL Utility program were applied for the modeling.