"Francois Bergeret" <[EMAIL PROTECTED]> wrote in av16dv$fbp$[EMAIL PROTECTED]:">news:av16dv$fbp$[EMAIL PROTECTED]:
> I have read on some references that the Wilcoxon test (or Kruskal > Wallis for more than 2 subgroups) has the assumption that the > distribution of values is symmetric aroiund the median. I'm not sure > this assumption is needed. In a book from Saporta there is a detail of > the test and I do not see this assumption. > > Is the assumption of symmetry needed for Wilcoxon or KW tests ? Please note that Wilcoxon developed two rank-based tests of central tendency: the rank-sum test (equivalent to the Mann-Whitney test) for independent samples and the signed-rank test for dependent samples. In this case I can assume you're referring to the former, since it's the one analogous to KW, but it's a good idea to specify which one you mean; a lot of people use "Wilcoxon test" to mean the signed-rank test. Whether that assumption is important or not depends on what hypothesis you're trying to test. If you're trying to test whether one distribution is stochastically higher than the other (i.e. one distribution is "slipped" relative to the other), then it's not important. If you're testing a hypothesis that's specifically about the difference between the group medians, then the assumption becomes important under some conditions. . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
