Comments on: Exploiting Hardy-Weinberg Equilibrium for association mapping

By: Thomas Mailund

Thomas Mailund — Sat, 26 Apr 2008 02:31:14 +0000

G: thanks for the reference, I’ll have a look at the paper.

By: Thomas Mailund

Thomas Mailund — Sat, 26 Apr 2008 02:29:22 +0000

Bob: They are not actually testing for HWE they just assume it. So yes, in a sense it is misspecification of the model. In our own experiments with it, though, we do not see much of a problem. We tried analysing the Wellcome trust data with permutated phenotypes (to get the null distribution) and we are not far from the expected null if we do that. Of course, it only works as a rough filtering of the markers, so (agian as they suggest in the paper) any hits should be analysed in more detail.

By: G

Thu, 24 Apr 2008 12:41:33 +0000

this slighlty older paper on HWE and disease models may also be of interest:
http://www.ajhg.org/retrieve/pii/S0002929707628948

By: Bob O'H

Bob O'H — Thu, 24 Apr 2008 08:24:42 +0000

The power of tests for HWE is notoriously bad, so in practice the gain in power is being traded off against model mis-specification. I guess I should read the paper to see how bad this is.