Figure 3 describes the onset of alcohol use over time by the four combinations of conduct problems and prior THC use described earlier. As expected, the survival rate (i.e. not having drank alcohol) was highest for offspring with no prior THC use and <2 conduct problems (39% of subgroup). The survival rate decreased to ~30% for the subgroups with one of these risk factors present. Most notably, the 12 offspring with both prior THC use and at least 2 conduct problems all had an alcohol onset by age 16, potentially indicating that the effect of THC use on the hazard function of drinking onset occurred primarily in offspring with conduct problems. The rates of alcohol use onset by age 22 across groups were significantly different (F(3,143)=3.04, p=.03), and post-hoc comparisons using the Games-Howell procedure (due to unequal group sizes) found significant group contrasts between the highest risk group and the two lowest risk groups only. While one potential reason for these group differences could be age heterogeneity across the subgroups, post-hoc analyses found that these groups were not significantly different in terms of offspring age at time of interview (Welch F(3)=1.41, p=.26).
