Although not central to our hypotheses, we note several other analysis results (results available from first author by request). The outcomes of past-year AUD and internalizing disorder, which were allowed to correlate in this model, were not significantly associated (r = 0.007 and r = −0.010 in the unlabeled and labeled groups, respectively). In addition, for descriptive purposes we conducted a supplemental regression of PAS on the internalizing disorder status variable. Those with persistent internalizing disorder had significantly higher PAS than those who were unaffected (standardized B = 0.192, p = 0.001), but differences in PAS with the remitted and incident groups were not statistically significant. Last, we compared the fit of our hypothesized model against a model where past-year AUD and internalizing disorder were independent variables, SNI and ISS were mediators, and PAS was the dependent variable. Our theorized model had a significantly better fit to the data than this alternative model, χ2(4) = 118.2, p < 0.0001.
