Non-parametric hypothesis tests

[Resa Pfu]

Hi there,

I have used non-parametric hypothesis tests (Mann-Whitney-U tests, Kruskall Wallis Test) to determine differences between groups regarding my non-normally distributed continous outcome variable (checklist score) in my thesis. Those examples of my codes:

ranksum checklistscore, by (device_group) porder exact

kwallis checklistscore, by (country)

This way the Mann Whitney-U test is a two-sided test (right?) and I am wondering about the Kruskal Wallis test, is it also a two-sided test?

Thanks for any help or clarifications

BW Resa

[Rich Goldstein]

each of these gives a 2-sided test

[Carlo Lazzaro]

Resa:
you may also want to consider:
1) a -bootstrap- -ttest-;
2) -regress- with -bootstrap- standard errors if you are going to compare the mean of =>3 samples.