Multiple Comparison With a Control for Multivariate Normal Means Based on Holm's Method

概要

論文の詳細を見る
In this study we consider the multiple comparisons with a control for multivariate normal means under the assumption that the common covariance matrix is unknown. Since it is difficult to derive the joint distribution of several related Hotelling's T^2-statistics which is necessary for determining the critical value for pairwise comparisons satisfying a specified significance level, we should use approximate approaches. Here we focus on a step down procedure called by Holm's method. Its critical value at each step is easily determined by Bonferroni's inequality. We compare Holm's method with two kinds of single step procedures having critical values determined by Monte Carlo simulation and Bonferroni's inequality respectively in terms of power of the test through the numerical examples.
東海大学の論文