Huh. I am sure there is one study you are missing. I remember another gwas right after the first.

I recall 4 gwass including latest.

You must be thinking about this: http://www.pnas.org/content/111/38/13790.abstract

I will include it too, then, although I am not sure about the validity of the proxy-phenotype method.

How many of the 7 alleles have other ones in LD?

At some point, maybe you should try weighting the factor analysis as some "hits" will be more reliable than others.

(Could you make a table for us? -- listing x independent alleles along with the ones in LD.)

I will add it to the paper I am working on. Gonna post it here soon.

Huh. I am sure there is one study you are missing. I remember another gwas right after the first.

I recall 4 gwass including latest.

You must be thinking about this: http://www.pnas.org/content/111/38/13790.abstract

I will include it too, then, although I am not sure about the validity of the proxy-phenotype method.

How many of the 7 alleles have other ones in LD?

At some point, maybe you should try weighting the factor analysis as some "hits" will be more reliable than others.

(Could you make a table for us? -- listing x independent alleles along with the ones in LD.)

The alleles were 9 in total, because Zoidberg signalled a GWAS I had forgotten about and it included 3 hits but one was in LD with another (which in turn was in LD with another one: they were in LD with each other, so this constitutes a genuinely replicated hit).
You can find my paper published here: http://dx.doi.org/10.6084/m9.figshare.1393160

Huh. I am sure there is one study you are missing. I remember another gwas right after the first.

I recall 4 gwass including latest.

You must be thinking about this: http://www.pnas.org/content/111/38/13790.abstract

I will include it too, then, although I am not sure about the validity of the proxy-phenotype method.

How many of the 7 alleles have other ones in LD?

At some point, maybe you should try weighting the factor analysis as some "hits" will be more reliable than others.

(Could you make a table for us? -- listing x independent alleles along with the ones in LD.)

The alleles were 9 in total, because Zoidberg signalled a GWAS I had forgotten about and it included 3 hits but one was in LD with another (which in turn was in LD with another one: they were in LD with each other, so this constitutes a genuinely replicated hit).
You can find my paper published here: http://dx.doi.org/10.6084/m9.figshare.1393160

I don't understand table 2. Your 9 allele has zero correlation with the 4 allele and a negative correlation with NIQs (but you report a 0.8 correlation). You have Han at -1.58 and South Han at 0.350 and Gambian at 1.2. Something looks screwy. Maybe I'm reading it wrong.

Huh. I am sure there is one study you are missing. I remember another gwas right after the first.

I recall 4 gwass including latest.

You must be thinking about this: http://www.pnas.org/content/111/38/13790.abstract

I will include it too, then, although I am not sure about the validity of the proxy-phenotype method.

How many of the 7 alleles have other ones in LD?

At some point, maybe you should try weighting the factor analysis as some "hits" will be more reliable than others.

(Could you make a table for us? -- listing x independent alleles along with the ones in LD.)

The alleles were 9 in total, because Zoidberg signalled a GWAS I had forgotten about and it included 3 hits but one was in LD with another (which in turn was in LD with another one: they were in LD with each other, so this constitutes a genuinely replicated hit).
You can find my paper published here: http://dx.doi.org/10.6084/m9.figshare.1393160

I don't understand table 2. Your 9 allele has zero correlation with the 4 allele and a negative correlation with NIQs (but you report a 0.8 correlation). You have Han at -1.58 and South Han at 0.350 and Gambian at 1.2. Something looks screwy. Maybe I'm reading it wrong.

Oh dear, you are absolutely right. I had some vectors in the alphabetical order and other grouped by continents. Fortunately this didn't compromise the analysis as the problem was only one of transcription from Excel to Word and it is not in the raw data set. Thanks for pointing this out! I have now uploaded a new manuscript with the correct values!

Oh dear, you are absolutely right. I had some vectors in the alphabetical order and other grouped by continents. Fortunately this didn't compromise the analysis as the problem was only one of transcription from Excel to Word and it is not in the raw data set. Thanks for pointing this out! I have now uploaded a new manuscript with the correct values!

What about the list of SNPs with those in LD? By the way, what's the rationale for excluding those in LD when it comes to the averaging of frequencies (not factor analysis).

Oh dear, you are absolutely right. I had some vectors in the alphabetical order and other grouped by continents. Fortunately this didn't compromise the analysis as the problem was only one of transcription from Excel to Word and it is not in the raw data set. Thanks for pointing this out! I have now uploaded a new manuscript with the correct values!

What about the list of SNPs with those in LD? By the way, what's the rationale for excluding those in LD when it comes to the averaging of frequencies (not factor analysis).

Also, is this paper citable? You're not planning any substantive changes are you?

Also, is this paper citable? You're not planning any substantive changes are you?

Also Lynn and Cheng (2013) give a Dai intra-national IQ of 94. Paper attached. Though, I think that it's a bad idea to use both intra-national and international IQs in the same analysis. There are clearly population level effects between countries.

Oh dear, you are absolutely right. I had some vectors in the alphabetical order and other grouped by continents. Fortunately this didn't compromise the analysis as the problem was only one of transcription from Excel to Word and it is not in the raw data set. Thanks for pointing this out! I have now uploaded a new manuscript with the correct values!

What about the list of SNPs with those in LD? By the way, what's the rationale for excluding those in LD when it comes to the averaging of frequencies (not factor analysis).

Well there are some very small variations in frequency for the two I checked using that genome browser. I couldn't get one in LD mentioned in the Duxide's paper. It showed an error in the genome browser. Would like to know about the others. Will take a look and see.

I checked 11 in LD from that Charge study and they are very similar in frequency but with a few relatively large changes with highest around 15% for a few populations.

You can use these 2 websites to check them using the sup table from the charge study.
Genome browser:
http://popgen.uchicago.edu/ggv/
SNP nexus to get the correct chromosome. The ones from the sup do not work. So you have to copy the snp then do the search and get the chromosome position and number.
http://www.snp-nexus.org/

Sup table.
https://sheet.zoho.com/editor.do?doc=b496449cf5bf9d3db9bf0b12fb011739694cf9ad7678490b18d8e9cca8cf599f87f5b68514ed1265c727e672c57e64d64c2c6693d13fad1a23c7f4b34bb0254d

So yeah I'm not sure exactly how the LD thing works here or weather they are actually the same gene or not but they are similar in frequency for the most part(by population).

Oh dear, you are absolutely right. I had some vectors in the alphabetical order and other grouped by continents. Fortunately this didn't compromise the analysis as the problem was only one of transcription from Excel to Word and it is not in the raw data set. Thanks for pointing this out! I have now uploaded a new manuscript with the correct values!

What about the list of SNPs with those in LD? By the way, what's the rationale for excluding those in LD when it comes to the averaging of frequencies (not factor analysis).

The SNPs in LD are reported at the bottom of table 1(asterisk).I included only the hits with genome wide significance (p= P ≤ 5 × 10(-8)). The rationale for excluding those in LD is that they are not indpendent observations. Using them to calculate average frequencies would give undue weight to the SNPs in LD...that is to say they would skew the average towards them, treating them as 3 when in fact they represent only one genetic locus. One thing I could have done would have been to average the frequencies of the 3 SNPs in LD and use the average frequency of them instead of that of 1 of the 3 SNPs, and treat this as a single SNP (and not as 3 separate SNPs). This would have given a slightly more accurate result but the frequencies are so similar that I doubt it'd have altered the overall picture. However, this would have made inflated the factor loadings of that genetic locus compared to the others simply due to better accuracy and not just because the SNP is necessarily a stronger target of selection.

No major changes to this paper are planned so you can cite it as: Piffer, Davide (2015): A review of intelligence GWAS hits: their relationship to country IQ and the issue of spatial autocorrelation. Figshare.
http://dx.doi.org/10.6084/m9.figshare.1393160

To see if they are in LD, just correlate them. If they are in LD, then r>.95.

LD is based on a correlation between alleles within populations. It has nothing to do with correlation between alleles among populations. There are two ways to calculate LD. The first, quickest way is to compute the distance on the genome. Alleles within 500Kbp of each other are said to be in LD with each other. Alternatively, LD can be computed using a software run on 1000 Genomes vcf files. When r>0.8, they are in strong LD. Moderate LD is r>0.5. Of course alleles that are in strong LD tend to have very high inter-population correlations, too but not viceversa (most alleles with high cross-population correlation are not in LD).
E.g. using Vcftools:

c:/Users/Davide/vcftools/bin/vcftools --vcf ALL.chr1.phase3_shapeit2_mvncall_integrated_v5.20130502.genotypes.vcf --geno-r2 --out chr1_ld --snp rs11584700 --snp rs11588857

I reported the location of the 3 SNPs and it's evident that they are within 500Kb of each other: location: 98,572,120 (rs10457441); location: 98,584,733 (rs9320913);
location: 98,553,894 (rs1487441)

c:/Users/Davide/vcftools/bin/vcftools --vcf ALL.chr1.phase3_shapeit2_mvncall_integrated_v5.20130502.genotypes.vcf --geno-r2 --out chr1_ld --snp rs11584700 --snp rs11588857

I reported the location of the 3 SNPs and it's evident that they are within 500Kb of each other: location: 98,572,120 (rs10457441); location: 98,584,733 (rs9320913);
location: 98,553,894 (rs1487441)

Ok, I had imagined that the researchers determine that the hits, say for the 12 allele study, which were in LD represented independent effects. This can be done e.g., "Accommodating linkage disequilibrium in genetic-association analyses via ridge regression". I figured that you were using LD in an idiosyncratic way of i.e., not representing unique hits useable in FA (a sort of between population "LD"). Didn't read through the papers.

Ok, I didn't know that.

Was this a new hit?

http://www.researchgate.net/publication/276026039_Independent_evidence_for_an_association_between_general_cognitive_ability_and_a_genetic_locus_for_educational_attainment

Ok, I didn't know that.

Was this a new hit?

http://www.researchgate.net/publication/276026039_Independent_evidence_for_an_association_between_general_cognitive_ability_and_a_genetic_locus_for_educational_attainment

No, it is in LD with the same 3 hits that are in LD. So this is the third replication of that locus for ed attainment

Ok, I didn't know that.

Was this a new hit?

http://www.researchgate.net/publication/276026039_Independent_evidence_for_an_association_between_general_cognitive_ability_and_a_genetic_locus_for_educational_attainment

No, it is in LD with the same 3 hits that are in LD. So this is the third replication of that locus for ed attainment

So its pretty much just 1 allele that has been replicated?

Does anyone know the ethnicity of the independent samples they used?