Via Steve Sailer: Common genetic variants associated with cognitive performance identified using the proxy-phenotype method
Haven't got the time to read it today, but seems interesting.
Back to [Archive] Other discussions
Via Steve Sailer: Common genetic variants associated with cognitive performance identified using the proxy-phenotype method
Haven't got the time to read it today, but seems interesting.
Yes it's a further replication of the 3 SNPs we used.
Via Steve Sailer: Common genetic variants associated with cognitive performance identified using the proxy-phenotype method
Haven't got the time to read it today, but seems interesting.
Yes it's a further replication of the 3 SNPs we used.
Were any of the other SNP associations reasonably reliable (i.e., low probability of being a false positive) ?
Now that your 3+1 SNPs have been re-replicated on cog data (no false positives in this respect), the issue seems to be whether a larger set of relevant SNPs would show more or less the same global distribution. Is there a way to determine how many SNPs you would need for reasonably robust results. (Coop's standards are simply absurd, but I think you need more than 4 alleles to draw a reasonably confident conclusion -- just not sure how many. 25? 50? 100? 200? What percent of the variance do they need to explain?) (Obviously, this is more of an epistemic than scientific question.)
That depends on the issue of representativeness of the first SNPs. We know that the first ones to be found are the ones with the largest effect sizes (they are thus the first to appear out of the noise in the data). The question is really whether these first few SNPs had similar selection force applied to them compared with the remaining. I don't know. But if the results hold up recently when we have the first 20 SNPs, I'm going to call it and say the debate is over.
In a recent comment (last two months?) on WH, Harpending wrote that 20 would be strong proof and 50 an incontrovertible result (IIRC; finding it is too much of a bother using my cell phone).
In a recent comment (last two months?) on WH, Harpending wrote that 20 would be strong proof and 50 an incontrovertible result (IIRC; finding it is too much of a bother using my cell phone).
So it was said.
But 50 SNPs will probably only explain ~5% of the variance owing to genes. So you will have a number of critics that argue along the lines of Coop. Of course, I could care less about them. I'm just looking for a justifiable rule of thumb.
We're running a GWAS that will settle the issue forever. We cannot disclose the results now but they will be made public within weeks.
© Copyright 2024 OpenPsych - All Rights Reserved