ABSTRACT
We saw in Chapter 5 that general factors may be obtained using a maximum likelihood rather than a least squares criterion, and we called these canonical factors. It turns out that canonical factors (Rao, 1955; Harris, 1962), and yet another variety, alpha factors (Kaiser & Caffrey, 1965) can be obtained via the same basic eigenvalue-eigenvector procedure as principal factors, by rescaling the starting correlation or covariance matrix before carrying out the calculation.
