Re: st: non-parametric tests

[Ron�n Conroy <rconroy@rcsi.ie>]

On 17 Feabh 2006, at 15:45, ottone@unipmn.it wrote:

> Dear all,
> we ran an experiment where we asked people how many tasks they  
> wanted to
> carry out under two different contracts. We ran both the Wilcoxon  
> and the
> Sign tests on that data. I think that both tests check whether people
> prefer working more under one of the two contracts. Probably, the  
> Wilcoxon
> test takes the magnitude of the difference into account, while the  
> sign
> test does not. Am I right?
> If so, what does it mean if the difference between people’s effort in
> contract A and in contract B is significant when I run the Wilcoxon  
> test
> but not when I consider the Sign test? In my mind, the opposite  
> would be
> more plausible…What’s the matter with my conclusions?
> Thank you very much for your help

The sign test is a pretty hopless test that loses almost all the  
information in your data, reducing all differences to positive or  
negative. I'm not sure it's been used in real life in a while. Go  
with Wilcoxon.

The same applies to the queries in your other two emails.

Ronán Conroy
rconroy@rcsi.ie




*
*   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/