Multivariate analyses were conducted with logistic regression since the dependent variables were dichotomous. Initial models tested for family structure variation in the likelihoods of adolescent smoking, drinking, and co-occurring risk behaviors, controlling for adolescent characteristics (i.e., age, gender, and race-ethnicity). Subsequent models included maternal socialization, modeling, and family economic context measures to examine the role of family processes. All analyses were conducted in Stata with the survey procedures to account for the complex sampling design of the Add Health study.
