Nonparametric Analysis of Covariance with Ordered Alternatives
研究了在考虑协变量X时，如何用非参数方法比较k个总体在响应变量Y上的有序差异，提出了两种检验统计量，并通过模拟评估了小样本表现，最后给出了逐步多重检验方法。
SUMMARY The problem of comparing k populations with respect to a univariate response variable Y against an ordered alternative is considered when a covariate X is taken into account using the matched-observations method of Quade (1982). Two test statistics are considered: one is Quade's statistic generalized for arbitrary k and the second is the Terpstra-Jonckheere statistic generalized for grouped data. Some simulations are carried out to study their small sample performance for different distributions of Y and X and for different sizes of the matched-observations classes. Finally, a general method of devising stepwise multiple testing procedures with fixed experiment wise error is applied to our problem of nonparametric analysis of covariance with ordered alternatives. An example of a neighbouring plots model is reanalyzed.