The relationship between son’s problem behavior and parenting: comparing a GEE and fixed effects model

• In Event: Individual and Family Risk Factors: Impact on Development and Maintenance of Antisocial Behaviour

Abstract

This study investigates whether offspring problem behavior impacts parenting through transactional processes comparing two statistical models. Most previous studies have employed models such as Generalized Estimating Equations (GEE) to allow for the nested structure of the data, where several assessments within the same person are analyzed. GEE explores differences between as well as within individuals. Studies using GEE thus investigate whether children who show more problem behavior on average have parents with worse parenting techniques. Of keen interest is whether a change in child behavior is related to a change in parent behavior in the same parent-child pair. Using data from the Pittsburgh Youth Study we use fixed effects regression models (conditional logistic regression) to study within-individual (or in this case within-parent-child-pairs) change. Because most studies so far have used population averaged marginal models (Generalized Estimating Equations or GEE), we also run GEE models to compare results and investigate whether the correlation between child and parent behavior can be explained by differences between or within individuals.

Authors

• Sytske Besemer, University of California, Berkeley

• Dustin Pardini, University of Pittsburgh

• Rolf Loeber, University of Pittsburgh

• Stephen Hinshaw, University of California, Berkeley