The Minimal Transformation to Orthonormality

Abstract

The solution is presented to the problem of finding that orthonormal matrix closest in the least-square sense to a given matrix of full rank. An application is shown to the problem in multiple regression analysis of determining the “importance” of each independent variable.