Abstract

Multiple component analysis of multivariate data from sample collections classifies environments on the assumption that the number of distinguishable environments is about equal to the number of positively correlated groups of variables, and that two positively correlated variables reach maximum development in a similar environment. Each group is a 'component' and each sample is classified according to the component it most resembles. Other statistical methods restrict the number of samples to about 200 because of computer limitations, but the multiple component method restricts only the number of variables. Using Bahama Bank samples, the various statistical methods are illustrated and results are compared.

You do not currently have access to this article.