Effect of Additional Variables in Principal Component Analysis, Discriminant Analysis and Canonical Correlation Analysis

摘要

In this paper, the authors derived asmptotic distributions of changes in certain functions of the eigenvalues of the sample covariance matrix, MANOVA matrix and canonical correlation matrix when some variables are added to the original sets of variables. The above results are useful in finding out as to whether the new variables give additional information for statistical inference; multivariate analysis; Wishart distribution. (Author)