Abstract
A method is presented which enables one to obtain confidence intervals for the probability of misclassiflcation for a given discriminant function using the observations from which it is computed. It is similar to Smith''s [1947] method in that it re-uses all of the data, but dissimilar in that the estimates one obtains are unbiased for the probability of misclassification for a discriminant function based on n1 - 1 and n2 observations or n1 and n2[long dash]1 observations.

This publication has 1 reference indexed in Scilit: