ABSTRACT
As we said we use B0 as a classification statistic for the problem in Eq. (5.18). That is, if B" > 0 we choose category II 1• Otherwise, we choose category Il2. This classification rule has the following intuitive appeal. The relation Bn > 0 implies en(./11 ,g) > e0 (],,f). Because e0 (f,g) represents a kind of distance between f and g, the rule implies./,, is nearer to f than g. The associated misclassification probabilities of B":
are evaluated as follows.
