On the Performance of Some Multiple Comparison Procedures with Unequal Variances

This paper gives the results from a computer simulation study concerning the type I, II and III error rates as well correct decision rates for 5 methods for pairwise mean comparisons when the population variances are unequal. These are the GH, C, H2, T2 and T3 procedures as well as modifications of...

Full description

Saved in:
Bibliographic Details
Published in:Scandinavian journal of statistics 1982-01, Vol.9 (4), p.241-247
Main Author: Markku P. Korhonen
Format: Article
Language:English
Subjects:
Error rates
Estimation methods
False negative errors
Population estimates
Population mean
Sample size
Significance level
Statistical discrepancies
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This paper gives the results from a computer simulation study concerning the type I, II and III error rates as well correct decision rates for 5 methods for pairwise mean comparisons when the population variances are unequal. These are the GH, C, H2, T2 and T3 procedures as well as modifications of the three latter methods. The results suggest that a higher type I error rate does not necessarily mean a lower type II error rate, i.e. the selection of the "best" method should not be based on the type I error rates only. The results indicate also that the nominal level of significance may be exceeded in the H2, T2, T3 and GH methods when some but not all sample sizes are small. The same liberalism of the modified versions of the methods is also observed in the equal and small sample size case. When the smallest sample size is greater than 5 then only the GH method seems to be liberal.
ISSN:0303-6898
1467-9469