Finally, a series of supplemental analyses were performed to confirm the robustness of our findings. In brief, first, MPNI impulsivity-hyperactivity (hereafter, impulsivity) ratings analyses were run in a similar manner to the main aggression ratings analyses (mean differences of impulsivity between ASPD and non-ASPD cases, correlations between informants, univariate models with only impulsivity ratings), as well as residual models where the effect of impulsivity was removed from aggression and vice versa, and a test to see if adding impulsivity ratings into the multi-rater aggression model with the highest AUC value improved ASPD prediction. Second, an additional alternative model was explored where MPNI social anxiety ratings (instead of aggression) were used as the main predictor. Third, we modified the ASPD definition: endorsing at least one, two, or four behaviors, or removing sub-clinical ASPD cases (one or two behaviors) from analyses. Fourth, we created residual scores for direct and indirect aggression that removed the effect of one on the other. Finally, we examined attrition by comparing mean differences in age 12 parent and teacher ratings among all available FinnTwin12 participants and in
