Asymptotic Conservativeness and Efficiency of Kruskal-Wallis Test for K Dependent Samples
研究了K个相依样本下Kruskal-Wallis检验的渐近保守性，提出了检验多元分布对称性的广义Kruskal-Wallis检验，并比较了其与对齐秩检验的相对效率。
The robustness (asymptotic conservativeness) of Kruskal-Wallis test under certain departures from mutual independence of K univariate samples is established. This robustness provides a procedure for testing the equality of K marginal distribution functions on a broken random sample from a K-variate distribution which satisfies a mild condition. For the unbroken sample, a generalized Kruskal-Wallis test is proposed for testing the symmetry of a K-dimensional distribution function. The relative efficiency of the K-W test against the aligned rank order test is also examined under the normal shift model. (Author)