| Summary: | 碩士 === 淡江大學 === 數學學系 === 85 === The procedures of testing the equality of normal means inthe
conventional analysis of variance (ANOVA) are heavilybased on
the assumption of the equality of the errorvariances. Studies
have shown that the F-test is not robustunder the violation of
equal error variances. When thevariances are unknown and
unequal, Bishop and Dudewicz(1978)developed a design-oriented
two-stage procedure for ANOVA,which gives exact tests with power
and level independent ofthe unknown variances. In this paper we
propose a two-stagerange test procedure with tables for
implementation, to testthe null hypotheses in ANOVA models under
heteroscedasticity.Simulation results indicste that the power of
the new rangetest procedure is as good as the two-stage ANOVA
method. Anumerical example is also given.
|
|---|
