Multiple testing correction T-Tests

#1
Hello. I am running an unpaired t test for 20 different dependent variables comparing for 4 "groups" to each other. This means a lot of T tests. How do I best account for multiple testing corrections? I believe I should be using Holms method. But do I need to correct for the total of 20x4 T tests or just the four for each group comparison within each dependent variable? How will anything still be significant when accounting for 80 T tests? Wont my new alpha be incredibly low?
 

Karabiner

TS Contributor
#2
How will anything still be significant when accounting for 80 T tests?
120, if you want to compare each group with each other.
With 4 groups, pairwise comparisons means 6 possible pairings.
How will anything still be significant when accounting for 80 T tests? Wont my new alpha be incredibly low?
Yes, and it would deserve to be so.

Why do you have the approach to
a) carry out separate tests on 20 dependent variables, and
b) to compare groups pairwise via t-tests, instead of doing a global test beforehand, i.e. oneway analysis of variance?

Could you please describe the situation in more detail, i.e. research question(s), research design, variables, sample size (!)

With kind regards

Karabiner