AN EXPERIMENTAL COMPARISON OF STATISTICAL AND LINEAR PROGRAMMING APPROACHES TO THE DISCRIMINANT PROBLEM

摘要: This paper reports the results of an experimental comparison three linear programming approaches and Fisher procedure for discriminant problem. The include two formulations proposed by Freed Glover a newly mixed-integer, goal formulation. Ten test problems were generated each 36 cells in three-factor, full-factorial design. Each problem consisted 30-case estimation sample 1,000-case holdout sample. Experimental indicate that four was statistically preferable certain Sample-based rules are suggested selecting approach based on Hotelling's T2 Box's M statistics. Subject Areas: Statistical Techniques, Linear Models, Programming.