Before running logistic regression models of aggression with ASPD, we first established a base model with only sex and exact age at the time of ASPD assessment as predictors of ASPD. Following this analysis, initial aggression models were performed with only one main predictor—one informant’s total aggression scores—in the model at a time, adjusted for sex and age. In these logistic regression models, the aggression scores (total, direct, indirect) from each informant were standardized (mean = 0, SD = 1) to allow for direct comparison. Additionally, receiver operating characteristic (ROC) curves were generated for the models and the area under the ROC curve (AUC) was calculated with 95% confidence intervals (CI) to demonstrate the predictive utility of the different informants’ aggression ratings in ASPD prediction [32]. We also tested for possible sex interactions with the total aggression scores from the different informants.
