Abstract
We seek to elucidate the impact of social activity, physical activity and
functional health status (factors) on depressive symptoms (outcome) in the
China Health and Retirement Longitudinal Study (CHARLS), a multi-year study of
aging involving 20,000 participants 45 years of age and older. Although a
variety of statistical methods are available for analyzing longitudinal data,
modeling the dynamics within a complex system remains a difficult
methodological challenge. We develop an Autoregressive Structural Model (ASM)
to examine these factors on depressive symptoms while accounting for temporal
dependence. The ASM builds on the structural equation model and also consists
of two components: a measurement model that connects observations to latent
factors, and a structural model that delineates the mechanism among latent
factors. Our ASM further incorporates autoregressive dependence into both
components for repeated measurements. The results from applying the ASM to the
CHARLS data indicate that social and physical activity independently and
consistently mitigated depressive symptoms over the course of five years, by
mediating through functional health status.