ABSTRACT
This chapter concerns presenting the multivariate generalization of bivariate correlation and the multivariate generalization is called the canonical correlation. In canonical correlation, the relationship is examined by finding a linear composite of the Y variables and a linear composite of the X variables such that the scores derived from the Y composite are maximally correlated with the scores derived from the X composite. Although this definition of canonical correlation may sound very similar to multiple regression, there are some important differences between the two procedures. The goal of multiple regression analysis is to maximize the correlation between the criterion and set of predictors. In order to illustrate a multiple regression analysis with Statistical Analysis System (SAS), a simple example is presented with two predictors: the number of drivers and the maximum temperature in January, used to predict the number of auto fatalities in January across the 50 states.
