Listy Biometryczne - Biometrical Letters Vol. 35(1998), No. 2, 83-93
|
ON THE ROBUSTNESS OF MULTIPLE COMPARISON PROCEDURES
Daniel Rabczenko1, Wojciech Zieliński2 1National Institute of Hygiene, Department of Medical Statistics, Chocimska 29, PL-00-791 Warszawa 2Department of Mathematical Statistics and Experimentation, University of Agriculture, Rakowiecka 26/30, PL-02-568 Warszawa e-mail: daniel@galaxy.medstat.pl, wojtek.zielinski@omega.sggw.waw.pl |
The paper concerns investigations on the robustness against non-normality of multiple comparison procedures. The probability of obtaining a division of means which is compatible with the true division is used as the robustness criterion. Results of Monte Carlo experiments suggest that the probability of correct decision increases with the proportion of observations coming from the contaminating distribution.
multiple comparisons, simultaneous inference, ANOVA, robustness.
