This is an easier to read version of the Canonical Correlation Comparison Table of Table 6.1 than was possible within the constraints of printing.

This table shows only the canonical variate loadings from each of the four analyses, and the discrepancy scores between pairs of analyses, both by rows and also by columns. The upper left matrix of loadings is Whitley’s initial sample with the loadings for his replication sample right next to it. Below these two are the corresponding analyses for our SAS re-analyses of his initial and replication sample data.

Explanation of Red Entries

The polarity on the canonical correlation latent variables can (and often does) switch from positive to negative in a seemingly random fashion without affecting the essential results. The only adverse effect of such polarity shifts is in the spatial orientation for plotting and for comparisons. For that reason, we have “corrected” these arbitrary variations by reversing the polarity on any latent variable in the four sub-tables that didn’t have a positive value in the boxed “three large loadings” of each of the predictor variables. All switches in polarity are shown in the table with the entries being colored red. (See the third full paragraph on page 290 of the book for a similar explanation of the use of bold entries in Table 6.1 on page 291 for this same purpose.)