Re: Mann-Whitney

Posted by Marta Garcia-Granero on Nov 14, 2011; 7:24am
URL: http://spssx-discussion.165.s1.nabble.com/Mann-Whitney-tp4989468p4989991.html

El 13/11/2011 20:41, Susan M Seibold-Simpson escribió:

Hi all-

I have a student who is trying to correct for ties in a Mann-Whitney test in SPSS v 18. I am unsure as to how to advise her.  Can anyone help? Thank you, Sue

Hi Susan:

The asymptotic p-value given in the output is already corrected for ties. If your student wants to use the exact p-value because sample size size is low, then the "2*(Exact one-tailed significance)" given in the output is not corrected. If the Exact Tests module is installed, you can ask for an exact p-value that takes into account the observed tie pattern. There is a freeware program that computes it (Wilcox.exe, by Jerome Klotz, available at  http://pages.cs.wisc.edu/~klotz/ ), but the procedure is not straightforward, since the tie pattern (the exact sequence every tie happens when data are sorted in ascending order) must but known.

If you really need the tie-corrected exact p-value, I think can write some SPSS code that will give you all the input the freeware applications needs. Right now, I have an Excel application that does it.

HTH,
Marta García-Granero