Finally, we tested for group differences in hazard functions. Covariate effects are expressed in hazard ratios (HR; e.g., difference in log hazard initiation for boys vs. girls, or for youth who experienced a parental divorce/separation vs. did not). We also tested interactions between divorce/separation and perceived stress, family history of drinking problems, and parent drinking as well as interactions of each of these variables with sex in predicting drinking onset.
