抄録
Two types of measures(direct and indirect measures)are defined to assess the closeness between a given vector and an eigenvector of a certain positive definite matrix. By using some results of multivariate distribution theory it is shown that the indirect measure is the reasonable one. From the viewpoint of the closeness of eigenvectors we deal with the following two problems in principal component analysis: Does a given vector approximate the coefficient vector of some principal component?; How many principal components should be taken into account?
