A new proposal for calculating the optimal discriminant function

This paper was developed in the Centro de Sanidad Animal and the Department of Mathematics of Universidad Nacional Agricola in 1997. One of the most frequently used multivariate statistical technique was employed: Discriminant Analysis. The existing different discriminant rules were studied in two populations with continuous random variables, which are included in different statistical packages, for establishing a methodology for calculating the optimum linear discriminant function. The classification error for reclassification and crossvalidation methods are included. The analysis of the developed study is completed with two examples with real data.