To demonstrate the predictive validity of the CTF, we conducted regression analyses in SAS version 9.4 (SAS Institute, Inc. 2013); analyses of CT Study and OZ‐ALC GWAS sample data used robust variance estimators to adjust 95% confidence intervals for the presence of familial clustering. Logistic regression analyses were used to examine the CTF‐associated risk for outcomes, in models that controlled for gender. For each overall sample, the CTF scores were first standardized in SAS to have means of zero and standard deviations (SD) of one. Odds ratios thus calculate risks associated with a one‐unit (i.e., 1 SD) difference in CTF score. Multinomial logistic regression was similarly used to determine risk, controlling for gender, for the number of licit and illicit drug dependence diagnoses associated with the CTF score. Post hoc analyses were performed in each sample's data that: (i) included neglect and EA in the regression models; and (ii) examined the correlation of the CTF score with neglect and EA (see Supplementary Methods).
