[R] Post-Hoc tests for Friedman Test?

[David Winsemius]

On Sep 20, 2009, at 9:05 PM, jon at terraspark.com wrote:

> Hi there all,
>
> This is my first post to the list and I'll first say a few things:
> - R is great!
> - The archives of this list have helped me solve all of my problems/ 
> questions so far
> - I only know enough statistics "to be dangerous"
>
> I'm looking for a way to do post-hoc tests for the Friedman test. I  
> have a dataset from a within-subjects design with 5 conditions where  
> some of the dependent variables are ordinal, resulting from (summed)  
> likert-scaled questionnaire data.
>
> From what I've read, I could use a wilcox.test on pairs of  
> conditions and adjust the p level, but is there something in R that  
> does a better job/automates this.
>
> I've seen references to the npmc package but that doesn't seem to do  
> what I'm looking for, because it only accepts a data frame with two  
> columns - i.e. there's no way to specify grouping/subject identifiers.
>
> Thanks,

There is a worked example in the coin package for using a permutation  
test to examine differences after a Friedman test. The authors,  
Hothorn , Hornik , van de Wiel, and Zeileis, call this method the  
Wilcoxon-Nemenyi-McDonald-Thompson test and cite: Hollander & Wolfe  
(1999), page 295

http://finzi.psych.upenn.edu/R/library/coin/html/SymmetryTests.html

--  

David Winsemius, MD
Heritage Laboratories
West Hartford, CT