RT Journal Article
SR Electronic
T1 An Empirical Study of the Robustness of Analysis of Variance Procedures in the Presence of Commonly Encountered Data Problems
JF American Journal of Enology and Viticulture
JO Am J Enol Vitic.
FD American Society for Enology and Viticulture
SP 131
OP 134
DO 10.5344/ajev.1982.33.3.131
VO 33
IS 3
A1 Shu Geng
A1 Paul Schneeman
A1 Wen-Jan Wang
YR 1982
UL http://www.ajevonline.org/content/33/3/131.abstract
AB This paper presents the results of Monte Carlo (computer generated) experiments in which the usual assumptions of variance homogeneity and normality are methodically violated, and the efficiency of various common remedial measures (rank transformation, weighting, and arcsine transformation) are compared using the ordinary analysis of variance procedure. In general, the ordinary analysis of variance procedure performs quite well, but the rank transformation procedure can actually produce misleading results if performed in the presence of unequally skewed populations. Therefore, routine tests of these assumptions before performing an analysis of variance are not necessary.