[R] mann whitney u

[Greg Snow]

How current is the literature?  Is the more recent literature using Mann-Whitney because of inertia rather than best practice?

The Mann-Whitney/Wicoxon test is a special case of a permutation test that has a shortcut computation.  Fast computers were not available when these tests were developed and so having the shortcut was very valuable.  These days with fast computers, that is much less important.

Before using the Mann-Whitney/Wilcoxon test, you should ask a few questions (actually these questions are probably appropriate for many different tests):

1. Do I understand what the test statistic is measuring?
2. Do I understand what null and alternative hypotheses are really testing?

(note: if the word median came to mind while answering either of the above, then the answer is "no").

3. Is this test meaningful for my project?
4. Is this test interesting for my project?
5. Can I explain what this test statistic/hypothesis really is to a lay person?
6. Do I want to explain this to a lay person?

If the answers to all the questions are yes, then use the Mann-Whitney/Wilcoxon test, otherwise I would suggest doing a permutation test on a more meaningful statistic.

Hope this helps,  

-- 
Gregory (Greg) L. Snow Ph.D.
Statistical Data Center
Intermountain Healthcare
greg.snow at imail.org
801.408.8111


> -----Original Message-----
> From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-
> project.org] On Behalf Of Mcdonald, Grant
> Sent: Wednesday, August 26, 2009 5:19 AM
> To: r-help at R-project.org
> Subject: [R] mann whitney u
> 
> Dear Sir,
> 
> I am comparing two samples using wilcox.test in R.  Literature appears
> to describe mann whitney u test as the most appropriate test to use on
> my data.
> 
> is the wilcox.test function equivalent to mann-whitney u?  Is there a
> way to gain the U-value as apposed to the W-value in R?
> 
> Thank you
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-
> guide.html
> and provide commented, minimal, self-contained, reproducible code.