Statistical analyses employed SUDAAN software (Research Triangle Institute, 2001) to account for complex design characteristics. Gender differences in unadjusted proportions and means were assessed using t-tests. Multivariate logistic regression models assessed the odds of lifetime alcohol dependence with respect to lifetime psychopathology, controlling for sociodemographic characteristics, lifetime drug abuse and nicotine dependence, personality disorders other than ASPD, and any significant interactions between gender and these covariates. Gender interactions were tested within a single model containing men and women; gender-specific odds ratios (OR) were presented regardless of whether significantly different or not. The same approach was followed in multivariate linear and logistic regression models examining the associations between psychopathology and clinical presentation and course indicators among lifetime alcoholics (n=4,781). Some variables were topcoded to reduce skew: volume of ethanol intake to 14.4 ounces, largest quantity of drinks to 36, age at onset of dependence to 45 years and interval to onset of dependence to 25 years. This involved the uppermost 3% to 5% of reported values for these variables. Post-hoc t-tests were conducted to assess the significance of differences in ORs and beta parameters across type of psychopathology and different populations.
