So I worked with non-parametric data (more than 3 groups) so I did the Kruskal-Wallis test to compare these groups. To see which group significantly differs from another one I did Dunn's Multiple comparison test with graphpad.

You can choose between a comparison of all groups or just selected ones. If I select two groups, graphpad shows me a significant difference between them. If I compare all groups in this experiment, the group I mentioned before is not significant anymore.

I am now wondering if it is ok to compare just the two to have a significant result in this experiment but compare all groups with each other in another experiment (same experiment but with another transcriptionfactor)? Or do I have to compare the same groups in every experiment?

To give an example: first experiment with transcriptionfactor A with 4 groups. Here I select group 3 and 4 and see a significant difference. However, in the second experiment with transcriptionfactor B (same 4 groups) I compare all (1 with 2; 1 with 3; 1 with 4....)

thank you in advance,