OPTIMIZING THE CLASSIFICATION PERFORMANCE OF LOGISTIC-REGRESSION AND FISHER DISCRIMINANT ANALYSES

Logistic regression analysis (LRA) and Fisher's discriminant analysis (FDA) are two of the most popular methodologies for solving classifica tion problems involving a dichotomous class variable and two or more a ttributes. Like other suboptimal classification methodologies, neither LRA nor FDA explicitly maximizes percentage accuracy in classificatio n (PAC) for the training sample (the sample on which the model is base d). A heuristic is described that shows early promise of increasing th e PAC of suboptimal models. The heuristic involves refining the cutpoi nt used by the suboptimal model to classify observations. This refinem ent is accomplished by applying univariate optimal discriminant analys is (UniODA) to the predicted response function values obtained by usin g the suboptimal model for training data. The UniODA cutpoint, rather than the cutpoint of the suboptimal model, is then employed to classif y both training and validity (hold-out) data. UniODA-refinement of LRA models is demonstrated by using 12 examples reflecting a variety of s ubstantive areas, including psychology, education, geriatrics, medicin e, biology, marketing, and geology. The mean PAC of UniODA-refined LRA was greater than that for nonrefined LRA in both training and validit y analyses. UniODA-refined LRA yielded greater validity PAC than nonre fined LRA in 6 of the 12 examples and lower validity PAC in only 1 of the 12 examples (exactly consistent results emerged for FDA). Discussi on focuses on directions for future research.