Structured Regression Analyses of Life Course Processes: An Example Exploring How Maternal Depression in Early Childhood Affects Children’s Subsequent Internalizing Behavior

Abstract

One of the specific aims of life course epidemiology is to assess the explanatory utility of three general hypotheses, namely the critical period hypothesis, the accumulation of risk hypothesis, and the effect modification hypothesis. A structured regression approach to this problem is illustrated with data from an ongoing longitudinal study of children and their families established in Adelaide in 1998-2000. A series of nested models that correspond to the alternative life course hypotheses were fit in an investigation of the effects of maternal depressive states in early childhood on internalizing child behavior at 9.5 years. Both linear and logistic regression models were considered. The structured regression framework showed the accumulation hypothesis was most plausible for these data. The analyses also provided some evidence of a critical period for the effect of maternal depressive status around child age 2 years on internalizing behavior at 9.5 years. The findings of this study suggest that comparing a suite of nested models to a full model can be useful in attempting to disentangle life course processes.