PSPP-BUG: [bug #47041] Mann Whitney U test results are not correct

[Eren Can Aybek]

Follow-up Comment #2, bug #47041 (project pspp):

It seems that when I have only two groups to compare (e.g. male, female) there
is no problem. 

But when I have 2+ groups to compare (e.g. elementary, high school and
university) and I only want to compare two groups, mean rank and sum of ranks
are different as well as U, W and p values.

Since Kruskal-Wallis H test doesn't have post-hoc tests, I need to compare
each group via Mann-Whitney U.

You can download the sample data from
http://ab.olcme.net/pspp/02%20-%20Veriler/04%20-%20ANOVA.sav

I run:

NPAR TESTS
/M-W tutum by bolum(1,2)

And results are different with the results of other softwares.

Thank you.

    _______________________________________________________

Reply to this item at:

  <http://savannah.gnu.org/bugs/?47041>

_______________________________________________
  Message sent via/by Savannah
  http://savannah.gnu.org/