Experimental evaluation of the classificatory performance of mathematical programming approaches to the three-group discriminant problem: The case of small samples

Although there have been several journal articles on the classificatory performance of mathematical programming approaches to the two-group discriminant problem, there has been no simulation study on the classificatory performance of mathematical programming approaches to the multiple-group problem reported in the literature. This study reports the results of a simulation experiment on the classificatory performance of a single-function and a multiple-function mathematical programming model relative to that of the standard parametric procedures for the three-group problem with small training samples. The effect of second-order terms on the classificatory performance of the mathematical programming models for the three-group problem is also investigated. Furthermore, this study theoretically examines the range of parameter values of a multiple-function mathematical programming model for which its number of misclassifications in the training sample cannot exceed that of a single-function model.