[R] Fastest way to do HWE.exact test on 100K SNP data?

Ritwik Sinha ritwik.sinha at gmail.com
Tue Jun 6 07:01:23 CEST 2006

Hi Anna,

I am not really answering your question, but, here goes my (unsolicited)
suggestion. Some time back I did some simulations to see how the HWE tests
performed. In particular I compared the exact test and the chi squared test.
Look at the attached figure Rplots.ps. I saw that for null simulations with
40 individuals, the HWE chi squared test was reasonably close to the
(expected) uniform distribution. However this was obtained using the
function HWE.chisq(X, simulate.p.value=F), the default seemed to have some
issues. Hence my suggestion would be, if you feel comfortable, to replace
the exact test with a chi sq test, at least at the screening level. Once you
identify a set of SNPs with "small" p-values, you could follow them up with
the exact test.

Ritwik Sinha
Graduate Student
Epidemiology and Biostatistics
Case Western Reserve University

http://darwin.cwru.edu/~rsinha <http://darwin.cwru.edu/%7Ersinha>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Rplots.ps
Type: application/postscript
Size: 7129 bytes
Desc: not available
Url : https://stat.ethz.ch/pipermail/r-help/attachments/20060606/d1bf16a8/attachment.ps 

More information about the R-help mailing list