Changed back to 0.9.

Okidoki, fine. I do not see any blocks having an average over 0.8 or 0.9, but we will see.

Venter middling

1621732 1557184
try({print(fem)})
[,1] [,2]
[1,] 1621732 1007968
[2,] 1557184 1076572

try({fisher.test(fem, alternative="two.sided")})

Fisher's Exact Test for Count Data

data: fem
p-value < 2.2e-16
alternative hypothesis: true odds ratio is not equal to 1
95 percent confidence interval:
1.108443 1.116225
sample estimates:
odds ratio
1.112333


Watson middling

1618178 1557184
try({print(fem)})
[,1] [,2]
[1,] 1618178 1011382
[2,] 1557184 1076572

try({fisher.test(fem, alternative="two.sided")})

Fisher's Exact Test for Count Data

data: fem
p-value < 2.2e-16
alternative hypothesis: true odds ratio is not equal to 1
95 percent confidence interval:
1.102295 1.110056
sample estimates:
odds ratio
1.106149


Watson low

81983 82330
try({print(fem)})
[,1] [,2]
[1,] 81983 64309
[2,] 82330 64199

try({fisher.test(fem, alternative="two.sided")})

Fisher's Exact Test for Count Data

data: fem
p-value = 0.4255
alternative hypothesis: true odds ratio is not equal to 1
95 percent confidence interval:
0.9796666 1.0087081
sample estimates:
odds ratio
0.9940815


Venter low

82150 82330
try({print(fem)})
[,1] [,2]
[1,] 82150 64150
[2,] 82330 64199

try({fisher.test(fem, alternative="two.sided")})

Fisher's Exact Test for Count Data

data: fem
p-value = 0.8494
alternative hypothesis: true odds ratio is not equal to 1
95 percent confidence interval:
0.984084 1.013277
sample estimates:
odds ratio
0.9985758


Venter high

11213 10397
try({print(fem)})
[,1] [,2]
[1,] 11213 4379
[2,] 10397 5203

try({fisher.test(fem, alternative="two.sided")})

Fisher's Exact Test for Count Data

data: fem
p-value < 2.2e-16
alternative hypothesis: true odds ratio is not equal to 1
95 percent confidence interval:
1.220715 1.345173
sample estimates:
odds ratio
1.281403


Watson high

11193 10397
try({print(fem)})
[,1] [,2]
[1,] 11193 4401
[2,] 10397 5203

try({fisher.test(fem, alternative="two.sided")})

Fisher's Exact Test for Count Data

data: fem
p-value < 2.2e-16
alternative hypothesis: true odds ratio is not equal to 1
95 percent confidence interval:
1.212489 1.335961
sample estimates:
odds ratio
1.272724

It should be simple. I have computed the snp frequencies in 1kg and have computed the derived alleles. Where on the priority list should I put it?

Ps. I think we should only do the rs-numbered SNPs, the other variants are often local to just one population.

Computing the number of blocks for the reference populations will be time consuming to code and to run: I have to split the reference population into each person, compute the number of good/bad/medium blocks for each person in the reference population and then compare against the brainiacs.

When we just counted the number of alleles you could use an aggregate score which was much easier.

Do you think it will be okay if I include the whole reference population in each block instead of creating one block for each person in the reference population?

This will mean that I create blocks with the mean value of the reference panel alleles in the block, so that if there are 220 haplotypes in the reference population each beneficial /detrimental allele in the block has a frequency of beneficial count / 220 and detrimental count / 220.

The blocks won't be true haplotype blocks though, but is this okay? This takes care of LD and is much easier to program.

Hope my question is understandable.

But I do not see how to do this.

If we have a block with these snps (8 people):

Snpname beneficial detrimental:
Snp1 10 6
Snp2 8 8
Snp3 15 1
Snp4 7 9

What should I count this as? Two beneficial blocks? One of each? And how do I compute it?

Im taking the rest of the weekend off, but I can skip the linkage part. It isn't that much work if I can do all ref people at once like you said. I imagine that if we do not some reviewer is gonna quibble. And if we get the same odds ratio that is a confidence inducing result we should mention in the paper.

I can do the ancestral:derived first, no problem.

His longest comment is both interesting and intelligent ("...Now let’s assume that there are not only “IQ genes”, but also genes that do not affect intelligence directly, but incline people to use their intelligence for real-world problem solving, as opposed to theoretical and philosophical pursuits..."), but I do not think we should spend too much time on such details, a small note should suffice. What are your thoughts?

• Find the ancestral allele frequency for 1kg.
• Find the ancestral allele frequency for the three groups of SNPs