A METHOD FOR SELECTING BETWEEN LINEAR AND QUADRATIC CLASSIFICATION MODELS IN DISCRIMINANT-ANALYSIS

A method for comparing the cross-validated classification accuracies o f linear and quadratic classification rules is presented under varying data conditions for the k-group classification problem. With this met hod, separate-group as well as total-group proportions of correct clas sifications can be compared for the two rules. McNemar's test for cont rasting correlated proportions is used in the statistical comparisons of the separate-group and total-sample proportions. The method is illu strated with some real data sets.