one way ANOVA for rank data?

#1
I was recently reading up on nonparametric alternatives to one way ANOVA and came across a description of ANOVA on rank data (http://www.stat.sc.edu/~hitchcock/soil_KW_sasexample705.txt). I found the article describing this test in: Conover and Iman. Rank Transformations as a Bridge Between Parametric and Nonparametric Statistics. Journal of the American Statistical Association. 35(3): 124-9.

I am interested in this topic because I am helping a friend to analyze data from his study, which has 1 control group and 3 groups which were exposed to different treatments. Most of the data does not meet the assumptions for ANOVA, so I am looking into nonparametric alternatives. I had originally planned to do Kruskal-Wallis and then individual wilcoxon tests to determine which groups are different.

I thought the Kruskal-Wallis test was the appropriate non-parametric alternative to performing a one way ANOVA. My question: is one way ANOVA on rank data appropriate (or rather, an appropriate non-parametric technique)? And if so, are there situations where this would be more or less appropriate than performing a Kruskal-Wallis test?
 
Last edited: