Additionally, our power for subgroup-specific patterns in ASPD prediction (e.g., male vs. female) may have been reduced due to having 67 ASPD cases (among 1347 participants). This 5% prevalence, however, is among typical population prevalence rates [1, 2], and relatively low case numbers is the natural consequence of utilizing a population-based sample for studies of uncommon conditions. Another limitation concerning power is that Marcus, Lilienfeld, Edens, and Poythress [40] suggest that ASPD exists on a continuum and that dichotomization—ASPD versus non-ASPD—could lead to a reduction in power. However, our supplemental alternative ASPD cut-off analysis showed that adjusting the ASPD diagnostic cut-off did not significantly affect the strength of the odds ratios, though AUC values fluctuated somewhat between models. Additionally, removal of sub-clinical ASPD cases also did not significantly improve odds ratios or AUC values. Thus, the main model indications that increasing levels of aggression predict ASPD hold no matter how we dichotomize ASPD. Finally, we also acknowledge that supplemental analyses and results related to social anxiety may reflect having only two items, rather than definitive lack of social anxiety association with ASPD.
