Asymptotic theory for canonical correlation analysis (Q1301588)

The author first presents a short review of canonical correlations and their sample versions. The main result of the paper is the derivation of the asymptotic distribution of the sample canonical correlations and coefficients of the canonical variates under the assumption that the nonzero theoretical canonical correlations are distinct and the sample is drawn from a normal distribution. Asymptotic distributions are derived also for the reduced rank regression model in the case that one set of variables is considered as independent and the other set as dependent. The author points out that some results on asymptotic distributions of sample canonical correlations published previously are not correct and he shows exactly where such derivations must be improved.