Canonical correlation analysis as a general linear model: An effective instructional tool

This article demonstrates in a concrete fashion that canonical correlation analysis subsumes all other least-square statistical methods and that, because it is the general linear model, it can be an effective instructional tool for teaching research methodology. The use of this general linear model framework illustrates that all least-square procedures involve the application of weights to derive synthetic scores which are the focus of all analyses and that all least-square procedures are correlational and therefore yield a measure of effect size to aid in the interpretation of results. Statistical analyses performed with a familiar statistical package, SPSS, on a small heuristic data set is used in the demonstration of various concepts.