Post by John Kemp on Aug 31, 2017 0:19:34 GMT
Hi,
I would very much appreciate your thoughts regarding the use of GCTA Cojo (--cojo-slct) in conjunction with GWAS summary statistics derived from BOLT-LMM using the non-infinitesimal model. From my limited understanding the beta and se estimated by Bolt-LMM under the non-infinitesimal and infinitesimal models are identical, but the test statistics and p-values differ. In my case it appears that the non-infinitesimal model provides a substantial improvement in power and I would ideally like to carry this over when identifying independent trait associated variants. Would one need to adapt the method for use with summary statistics derived from non-infinitesimal model?
Kind regards
John K
I would very much appreciate your thoughts regarding the use of GCTA Cojo (--cojo-slct) in conjunction with GWAS summary statistics derived from BOLT-LMM using the non-infinitesimal model. From my limited understanding the beta and se estimated by Bolt-LMM under the non-infinitesimal and infinitesimal models are identical, but the test statistics and p-values differ. In my case it appears that the non-infinitesimal model provides a substantial improvement in power and I would ideally like to carry this over when identifying independent trait associated variants. Would one need to adapt the method for use with summary statistics derived from non-infinitesimal model?
Kind regards
John K