Statalist The Stata Listserver


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: RE: mann-whitney test


From   "Nick Cox" <n.j.cox@durham.ac.uk>
To   <statalist@hsphsun2.harvard.edu>
Subject   st: RE: mann-whitney test
Date   Thu, 4 May 2006 17:35:55 +0100

In essence, you have panel data. 
I cannot see that Mann-Whitney is 
appropriate for any comparison 
unless you throw away all the 
information on changes over time. 

Nick 
n.j.cox@durham.ac.uk 

ottone@unipmn.it
 
> Sorry, but I'm confused! I ran two treatments and  32 
> subjects (organised
> in group of 4) participated to each treatment (10 periods). This means
> that I have only 8 independent observations in each period for each
> treatment. How should be my dataset to run the mann-whitney test? If I
> consider 320 observations in each treatment, I cannot say 
> that they are
> independent. How should I perform the test in this case?
 
> > There is some misunderstanding here, I guess.
> >
> > Mann-Whitney requires at a minimum all
> > the individual ranks. It can't work on just
> > input of averages, or even medians.

> >> I have to run the mann-whitney test to check the difference
> >> in the trend
> >> between two treatments in a public good game experiment.
> >> This test has to be run on average contributions of groups.
> >> Does it mean
> >> that I should have in my dataset a single observation for
> >> each group that
> >> represents the average contribution of the group over all 
> the periods?

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index