ABSTRACT
Conventional index methodology is based on a multivariate statistical technique related to discriminant analysis. Related, but distinct, techniques have been recommended for choosing parental material for crop improvement programs. The related techniques include vector analysis, canonical analysis, and multivariate distance analysis. Canonical analysis is a multivariate statistical method for representing data in dimensions which give the greatest discrimination among groups of individuals. The basic data are transformed so that the first canonical variate shows the greatest ratio of among-group to within-group variation. The multivariate statistical methods used for choosing parents for a plant breeding program serve primarily to reduce the number of variables to one or two. Beyond this obvious simplification, the methods group naturally into those in which parents are chosen because their mid-parent value is as close as possible to the specified objective of the breeding program, and those where parents are chosen as to maximize the genotypic variation within the progeny population.
