Post-test for nonparametrics?

#1
Hi!
I need some help regarding nonparametric tests doubts that are just *killing* me :(:

1) Is it mandatory to use nonparametric tests when I'm dealing with ordinal-type, qualitative data, such as "a lot", "a few", "none"?

2) Are there other nonparametric post-tests, beside Dunn's, to do pairwise comparisons when I have more than 2 groups?

3) Is MINITAB able to perform any pairwise comparison when dealing with nonparametrics? Even Dunn's?

GraphPad Prism could perform Dunn's post-test and I loved it! However, I'm trying to expand my horizons and playing a bit with MINITAB.

Thank you very much for the help! All the best!:)
 
#2
1. I think so, although not absolutely sure.
2. Good tests are nemenyi test and steel-dwars both after non-parametric anova (ie Kruskal wallis). otherwise use U-test (Mann-Whitney), median test, Friedman test. They all depend on ranks. Unfortunately several of them are not included in many programs, but perhaps SAS and R there are macros available.
3. Nevr used minitab. :(

Zanello said:
Hi!
I need some help regarding nonparametric tests doubts that are just *killing* me :(:

1) Is it mandatory to use nonparametric tests when I'm dealing with ordinal-type, qualitative data, such as "a lot", "a few", "none"?

2) Are there other nonparametric post-tests, beside Dunn's, to do pairwise comparisons when I have more than 2 groups?

3) Is MINITAB able to perform any pairwise comparison when dealing with nonparametrics? Even Dunn's?

GraphPad Prism could perform Dunn's post-test and I loved it! However, I'm trying to expand my horizons and playing a bit with MINITAB.

Thank you very much for the help! All the best!:)
 
#3
Thanks!

Hobgoblin,

Thanks for the help!:) I'll try to find someone who works with any of those (SAS or R) and see if they can perform such tests.

It's a shame that nonparametric post-tests are so hard to find in statistic software...:(