ABSTRACT
In Chapter 3, we examined breaking down the association between two sets of variables using multivariate regression analysis. This is the appropriate technique if our interest is in prediction, and if we wish to focus our attention primarily on the individual variables (both predictors and dependent) rather than on linear combinations of the variables. Canonical correlation is another means of breaking down the association for two sets of variables, and is appropriate if the wish is to parsimoniously describe the number and nature of mutually independent relationships existing between the two sets. This is accomplished through the use of pairs of linear combinations that are uncorrelated.
