Testing mediation using multiple regression and structural equation modeling analyses in secondary data

Abstract

Mediation analysis in child and adolescent development research is possible using large secondary data sets. This article provides an overview of two statistical methods commonly used to test mediated effects in secondary analysis: multiple regression and structural equation modeling (SEM). Two empirical studies are presented to illustrate the respective circumstances in which the two methods are most useful. One study examines the mediated effect of parents' social capital on parent involvement in Head Start programs through parent-child bond. The other study assesses the mediating effects of structured routine activities, delinquent association, and prosocial belief on the relationship between religiosity and juvenile delinquency.