Multiple Comparisons - Question

Question

[areyoujokingme]

I'm setting up a statistical design and I want to do GWAS on a bunch of SNP data I've gotten access to. I want to test only the sites that have been nominated in the past initially, so that my multiple comparisons correction will not be too overwhelming. I thought about also including any haplotype regions for the included variants as well. Anyway, it will still be significantly less than genome-wide comparisons.

My question is... if I then proceed to do a genome-wide association study on the *same data*, will I have to correct for the comparisons made in my initial comparison of just the nominated variants *AND* all comparisons made in the GWAS? What do the statisticians say?

Answer 1

[Devon Ryan]

There's no point in correcting twice for comparisons, so no, don't bother correcting for the initial comparisons when doing the full-scale comparisons (you'll be making some of the same comparisons and will already be correcting for them).