An alternative method for ordered ANOVA under unequal variances
LE3 .A278 2016
Bachelor of Arts
Mathematics and Statistics
Mathematics & Statistics
The ordered one-way analysis of variance (ANOVA) has drawn a lot of attention. Its test procedure differs from the unordered one-way ANOVA as it needs to detect not only the difference in mean, but also whether the means follow monotonic restriction. The Welch trend (WT) test, proposed by Roth (1983) is an ordered test based upon the Welch unordered test by Welch (1953) which allows unequal variances across the treatments. However, it is pointed out that the WT test has an inated type one error relative to its nominal level. We propose an alternative test for ordered one-way ANOVA which employs the parametric bootstrap method similar to Krishnamoorthy et al. (2007) in order to have well controlled type-I error.
The author retains copyright in this thesis. Any substantial copying or any other actions that exceed fair dealing or other exceptions in the Copyright Act require the permission of the author.