Modified ANOVA for unequal variances

被引:25
|
作者
Lee, S [1 ]
Ahn, C [1 ]
机构
[1] Sejong Univ, Dept Appl Math, Seoul 143747, South Korea
关键词
ANOVA; unequal variances; Behrens-Fisher problem; Welch test;
D O I
10.1081/SAC-120023874
中图分类号
O21 [概率论与数理统计]; C8 [统计学];
学科分类号
020208 ; 070103 ; 0714 ;
摘要
The departures from, the equal variances assumption can have a serious negative effect on the F-test of ANOVA. By using a ratio of sample size and non-pooled sample variance as a weight, the statistic F is modified in order to be applicable without the assumption of equal variances, and its approximate distribution is derived. This new test is very sensitive to the weights. Using simulation, its sizes and powers with the various weights are observed, and the comparisons are made with other known tests.
引用
收藏
页码:987 / 1004
页数:18
相关论文
共 50 条