Differences in Longitudinal Trajectories between Groups - The Multi-Group Latent Growth Components Approach

Purpose: In this article, we propose a multi-group approach for analyzing
complex nonlinear longitudinal trajectories. Method: The approach is based on
the latent growth components approach (LGCA) that offers a flexible
framework for defining growth components and extends the same for the use
with multiple groups. The approach benefits from known advantages of the
LGCA and adds more capabilities from the multi-group framework, that is, (1)
it can flexibly include complex nonlinear growth components, (2) incorporate
a measurement model for the latent state variables and latent covariates, (3) it
can model differences in growth components based on categorical covariates,
and (4) treat covariates and group weights as fixed or stochastic. Results and
conclusions: We demonstrate the approach using data from the Health and
Retirement Study that includes individuals diagnosed with cancer. We analyze
trajectories in depressive symptoms before and after the cancer diagnosis with
respect to a subset of categorical covariates (i.e., groups). We further present
the open-source R package semnova that implements the proposed approach
and makes it conveniently accessible for applied researchers.