An evaluation of some nonparametric multiple comparison procedures by Monte Carlo methods |
| |
Authors: | F. A. Lin J. K. Haseman |
| |
Affiliation: | Biometry Branch , National Institute of Environmental Health Sciences , P. 0. Box 12233 Research Triangle Park, North Carolina , 27709 |
| |
Abstract: | Computer simulation techniques were employed to investigate the Type I and Type II error rates (experiment-wise and comparison-wise) of three nonparametric multiple comparison procedures. Three different underlying distributions were considered. It was found that the nonparametric analog to Fisher’s LSD (a Kruskal-Wallis test, followed by pairwise Mann-Whitney U tests if a significant overall effect is detected) appeared to be superior to the Nemenyi-Dunn and Steel-Dwass procedures, because of the extreme conservatism of these latter methods. |
| |
Keywords: | ranks simultaneous statistical inference non-parametric tests type I and II error rates |
|
|