Abstract
In this paper, an algorithm related to all possible regression and discriminant analysis involving less than r components out of k is presented. For this purpose the modified binary increment is proposed to produce all possible subsets of size less than r out of K={1, 2, ……, k}. Refering to these subsets produced, our algorithm is designed to perform calculations successively. Two kinds of improvements over our previously proposed algorithms are presented and examined, with an explicit recommendation for the use of computers in the regression and discriminant analyses.
