“…There are general least-squares programs capable of reduced-model tests [e.g., SAS's GLM (Barr et al, 1976); Bock, 1965;BMDX63/11V (Dixon, 1973);Finn, 1972;Woodward & Overall, 1974]. These are not labeled as canonical correlation programs, since they work with sums of squares and cross-products (SP) matrices, rather than with correlation matrices; the procedures are nevertheless equivalent.…”