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

The criterion group classification scheme commonly used in conjunction with discriminant function analysis places individuals in the group for which their membership probability is the highest. This optimizes the proportion of correctly classified cases. The clinical utility of a discriminant based scheme may, however, be underestimated if the proportion of correctly classified cases is the only concern. For clinical purposes, attention must also be paid to minimizing serious diagnostic errors and to improving predictions for individuals. Clinicians can address these concerns by considering membership probabilities for each criterion group of interest. The utility of this approach is illustrated using data from a study in which discriminant analysis was used to identify the characteristics of problem drinkers who abstained, controlled their drinking, or continued to drink heavily after treatment.