Discriminant analysis via mathematical programming: Certain problems and their causes

In the past several years there has been some interest shown in the use of mathematical programming (i.e. specifically either linear programming or linear goal programming) for use in discriminant analysis-an area previously approached only by the more conventional methods of statistical analysis. However, those who have actually attempted to apply the mathematical programming approach have, in some instances, encountered a number of problems which raise questions about the usefulness of the methodology in such an application. In this paper, the authors describe these problems, identify their causes and suggest ways in which such problems may be avoided. This work represents an ongoing effort on the part of the authors in the development of an efficient revised approach to discriminant analysis via mathematical programming.