Whole experiment or groupwise p-value adjustments?

19 Views Asked by At

I have analyzed in an experiment a lot of groups and parameters. The parameter ranges are quite different:

Parameter 1: 0 to 1.2 Parameter 2: 0 to 1000 Parameter 3: -700 to 42 ... Parameter n: 7 - 120

My plan is to test with Kruskal-Wallis, whether the groups are different within a parameter and adjust the p-values with for example Bonforroni-Holm. In a second step I planned to compare the groups within a parameter and adjust the p-values ...?

Is it OK to adjust within are parameter (3 comparisons) or do I have to adjust to all comparison in this experiment (3 comparison * n)?

Thank you in advanced for your help