极值分布的方差分析检验

本文考虑位置参数和尺度参数未知时极值分布的拟合优度检验。所提统计量W_G类似于Shapiro和Wilk(1965)的正态性检验W,对n=3(1)25给出了W_G的系数表及分布的分位点表,还将W_G与Cramer-Von Mises型统计量A~2,W~2,U~2作了检验功效比较,指出W_G是关于极值分布拟合优度检验的较好统计量。

AN ANALYSIS OF VARIANCE TEST FOR THE EXTREME VALUE DISTRIBUTION

In the present paper we consider a goodness of fit test for the extreme value distribution when the location and scale parameters are both unknown. The proposed statistic W_G has analogy with Shapiro & wilk (1965) W test for normality. The coefficients and percentage points of W_G are tabled for n=3(1)25. We also make a comparison of powers between W_G and Cramer-Von Mises type statistics A~2, W~2, U~2 and point out that W_G is a good statistic for goodness of fit test for the extreme value distribution.