Optimal Bounds for the Distributions of Some Test Criteria for Tests of Dimensionality
研究了E和H服从相同协方差矩阵的Wishart分布时，HE−1的m-r个最小特征根的两个函数分布的最优上界，并用这些界评估维度检验中卡方近似的效果。
Optimal upper bounds are obtained for the distributions of two functions of the m - r smallest latent roots of HE−1, where E and H have Wishart distributions with identical covariance matrices; E has a central distribution while H has a noncentral distribution with unknown noncentrality matrix Δ of rank r. These bounds are then used to investigate the chi-squared approximation for some test criteria used in tests of dimensionality.