ABSTRACT
The term discrimination (in a nonpejorative statistical sense) can refer to the task of separating groups through linear combinations of variables maximizing a criterion, such as an F -ratio. The linear combinations themselves are commonly called Fisher’s linear discriminant functions. The related term classification refers to the task of allocating observations to existing groups, typically to minimize the cost and/or probability of misclassification. These two topics are intertwined, but here we briefly comment on only the topic of classification. Any applied multivariate analysis course should treat these two topics in much greater depth.
