Interpretations of Multiple Regression Borrowed From Factor Analysis and Canonical Correlation

Abstract
Because multiple regression (MR) is a simplified case of canonical correlation, the authors present an alternative interpretation of MR in which loadings, in addition to weights, may be computed for predictors. The loadings, computed by both canonical correlation and factor analysis, represent the correlations of each predictor with the predicted criterion score. The authors concluded that in some situations, researchers should examine those values in addition to testing for the significance of beta weights.