Kruskal Wallis - Specific multiple comparisons

[peterp]

Dear all,
I'm new to the board, but I hope you're able to help me.
I'm using SPSS for my statistics. My outcome is scale variables (eg height, weight, etc.) - a total of 20 different variables. I have 8 different groups where I need to test 8 relevant combinations per variable.
My first approach was to use the Mann Whitney U test (my sample size is small so I do not assume my data is normal distributed), however this does not account for the multiple comparisons problem.
Therefore I used the Kruskal Wallis test with multiple comparisons using Dunns and a Bonferroni correction. The problem is that I have 8 different groups, the pairwise comparison gives me 28 comparisons (all possible combinations) and corrects for 28 comparisons.
As mentioned, only 8 specific comparisons are relevant for me. This means that I underestimate the number of significant differences, since I correct for 28 comparisons.
Is there any way I can specify which group combinations I need and still correct for multiple comparions? Maybe another test or maybe a setting I can enable?

If anything is unclear, please tell me and I will try to clarify

Thanks in advance

Best regards
Peter

 

[Karabiner]

You can perform 8 single U-tests a spost-hoc tests.
Bonferroni-adjustment (p/8) you then can do by hand.

With kind regards

K.

 

[peterp]

Thank you for your response.
My alpha level is 0.05 which means that a corrected alpha level of 0.05/8 = 0.00625 should be considered significant.
In one case (using the two methods descriped in the original post) the U-test gives me a p-value = 0.004, however when corrected using Kruskal-Wallis (With Dunns procedure and Bonferroni correction in SPSS) the adjusted p-value is 0.020.
Is it possible to report these adjusted values instead of considering p-values below 0.00625 significant?

Thank you