Discrimination by means of components that are orthogonal in the data space

摘要

Drzanowski (J. Chemometrics, 9, 509 (1995)) proposed a method for obtaining so-called orthogonal canonical variates (henceforth called components) for discrimination purposes. In contrast with ordinary discriminant analysis, this method employs components that are orthogonal in the original data space. These components are derived in a successive way, thus optimizing discrimination of a component give the previously extracted components. Two alternative procedures are proposed to of a component given the previously extracted simultaneously, yielding a better overall discrimination.