To better visualize this interaction effect, a series of tables (not shown) of the onset of alcohol use from the earliest age of drinking onset in this sample (12 years) to age 22 was created using dichotomized subgroups of the main effect variables. For these analyses, prior THC use was still treated as a binary category (15% yes vs. 85% no), while conduct problems were dichotomized based on <2 problems (70%) vs. 2+ problems (30%). The cutoff of two conduct problems was selected because 28% of the sample reported one conduct problem, most commonly fights (33%), skipping school (28%), and stayed out/sneaked out of the house at night (27%), any one of which alone may only be marginally indicative of a significant behavior problem. For this interaction, four subgroups were created from the variables described above, with the survival function then calculated for each subgroup individually and graphed over time.
