Discriminant analysis and the selection of patients for controlled drinking programs: a methodological note