This is just a small question, so I can keep track.
I want to know specifically from the most recent version of that allele correlation study. From Piffer IIRC? The one where South Americans and South Asians had a similar polygenic score as SS Africans.
I assume its around 90?
Back to [Archive] Other discussions
In my opinion, the current estimates based on SNPs are not so reliable yet. However, when the new hits come out 'soon' (about 70-80 SNPs), we should be able to get a more reliable estimate. Personally, I'd guess around 80. This would fit the the general value of 85 for African Americans (with 20% European admixture). Different values are possible, but then one will have to introduce more complex explanations. For instance, suppose SSA genotypic IQ is about 90. Why then is the African American IQ only 85, including the expected boost from 20% European? Well, perhaps there was negative selection for cognitive ability among African Americans that lowered their genotypic IQ, or positive selection in Africa since they left. After all, the environment in SSA Africa (except for South Africa, the country) has been consistently horrible since they left while it was somewhat improved in the US.
I'll ask Piffer to reply here.
Perhaps the best thing to do is for the main scientists in the area to give predictions in the form of probability distributions (in the spirit of http://gjopen.com/). Mine would be something like N(80, 3) but I would have a fatter left tail.
An easier method to estimate the genotypic (SSA) African mean is to get one large sample that has both racial admixture and cognitive ability for some two-way admixtured population. African Americans would fit the bill. There are by now many such large samples that have socioeconomic data and they generally point in the same direction. Unfortunately (for science), they are not publicly available and often involve three-way splits (European, African, Amerindian) which make the estimation of a single group's genotypic mean more difficult.
Suppose we had a large dataset (e.g. N=5000) of African Americans living in the US, that we have their WAIS scores and that we are able to estimation racial admixture reliably (e.g. using 100k markers). Suppose that in this sample we observe a correlation of -.25 between African admixture and cognitive ability. Suppose the African Americans obtained a mean IQ of 85, have 80% African admixture and have an African admixture SD of 15. In this case, we calculate that to get to the African genotypic mean, one needs to go 20/15 = 1.33 SD towards more African admixture. We then multiply this by the African admixture x cognitive ability to to get 1.33*-.25 = -.3325. This is in standardized units, so we multiply by the IQ SD to get -.3325*15 = -4.99. Finally, we add this to the observed mean score, 85, to get 80.01. One can do the same exercise from the perspective of European ancestry. This again requires the ancestry x cognitive ability correlation, the ancestry mean and the ancestry SD.
In our (with John Fuerst) large meta-analysis of racial admixture x socioeconomic outcomes (https://osf.io/ydc3f/), we found an overall (across all samples, countries etc.) correlation of -.10. This is without doubt an underestimate for many reasons (lower admixture SD in many samples, poor S measures, substantial measurement error in the admixture estimates, the use of ordinal measures with few levels). Unfortunately, there is no good estimate of individual S x African admixture for a large US sample of African Americans. The best is probably Qi et al. (2012) who found a neighborhood SES based correlation of -.20 (N≈10k). I think this is likely an overestimate due to neighborhood amplification effects, so perhaps a true value would be around -.15. If we assume so and also assume a simple cognitive genetic admixture (difference between groups entirely due to genetic admixture related cognitive ability), then this would imply a cognitive ability x S correlation of -.15/-.25 = .60, which is roughly what one finds in the cognitive sociology literature.
So the numbers roughly fit with a genotypic mean of 80 for SSA, assuming no difference between African SSA ancestry and US SSA ancestry.
I'll ask Piffer to reply here.
Perhaps the best thing to do is for the main scientists in the area to give predictions in the form of probability distributions (in the spirit of http://gjopen.com/). Mine would be something like N(80, 3) but I would have a fatter left tail.
An easier method to estimate the genotypic (SSA) African mean is to get one large sample that has both racial admixture and cognitive ability for some two-way admixtured population. African Americans would fit the bill. There are by now many such large samples that have socioeconomic data and they generally point in the same direction. Unfortunately (for science), they are not publicly available and often involve three-way splits (European, African, Amerindian) which make the estimation of a single group's genotypic mean more difficult.
Suppose we had a large dataset (e.g. N=5000) of African Americans living in the US, that we have their WAIS scores and that we are able to estimation racial admixture reliably (e.g. using 100k markers). Suppose that in this sample we observe a correlation of -.25 between African admixture and cognitive ability. Suppose the African Americans obtained a mean IQ of 85, have 80% African admixture and have an African admixture SD of 15. In this case, we calculate that to get to the African genotypic mean, one needs to go 20/15 = 1.33 SD towards more African admixture. We then multiply this by the African admixture x cognitive ability to to get 1.33*-.25 = -.3325. This is in standardized units, so we multiply by the IQ SD to get -.3325*15 = -4.99. Finally, we add this to the observed mean score, 85, to get 80.01. One can do the same exercise from the perspective of European ancestry. This again requires the ancestry x cognitive ability correlation, the ancestry mean and the ancestry SD.
In our (with John Fuerst) large meta-analysis of racial admixture x socioeconomic outcomes (https://osf.io/ydc3f/), we found an overall (across all samples, countries etc.) correlation of -.10. This is without doubt an underestimate for many reasons (lower admixture SD in many samples, poor S measures, substantial measurement error in the admixture estimates, the use of ordinal measures with few levels). Unfortunately, there is no good estimate of individual S x African admixture for a large US sample of African Americans. The best is probably Qi et al. (2012) who found a neighborhood SES based correlation of -.20 (N≈10k). I think this is likely an overestimate due to neighborhood amplification effects, so perhaps a true value would be around -.15. If we assume so and also assume a simple cognitive genetic admixture (difference between groups entirely due to genetic admixture related cognitive ability), then this would imply a cognitive ability x S correlation of -.15/-.25 = .60, which is roughly what one finds in the cognitive sociology literature.
So the numbers roughly fit with a genotypic mean of 80 for SSA, assuming no difference between African SSA ancestry and US SSA ancestry.
This is just a small question, so I can keep track.
I want to know specifically from the most recent version of that allele correlation study. From Piffer IIRC? The one where South Americans and South Asians had a similar polygenic score as SS Africans.
I assume its around 90?
You are right. I recently issued a new version of the polygenic selection study. This included a new procedure that dealt with GWAS population-related biases in derived allele frequencies. Results weren't much affected (r= 0.97 between the two estimates). The other novelty is that I used more SNPs (100+ from Rietveld and others related to fluid intelligence from Davies et al.). Well, there was an interesting twist to the story as East Asians lost the top place, which was taken by Euros. Food for thought? Maybe, or just junk food until we get more g-related SNPs. It's entirely possible that the East Asian advantage is not related to general intelligence but rather to higher spatial abilities, which would account for their advantage in Mathematics but their relatively poor verbal abilities. Can someone find studies where this East Asian spatial>verbal tilt is taken into account when estimating national IQs? Also, it's been noted (again, reference lacking from my mind) that g as a construct isn't as robust among East Asians. Norms for Raven's progressive Matrices show that East Asians and Whites perform about equally well, followed by Hispanics (B=-0.15) and Black (B=-0.26):
http://www.eyeonsociety.co.uk/resources/RPMChangeAndStability.pdf
Be that as it may, here are the rankings that I estimated based on all the significant or near-significant GWAS hits for educational attainment and fluid intelligence (about 190 SNPs). These are polygenic scores corrected for derived allele status but the non-corrected rankings are virtually identical. Z scores are reported in descending order (from larger to smaller).
Toscani, Italy 1.558
Iberian, Spain 1.442
Finland 1.375
British, GB 1.205
Vietnam 1.046
Utah Whites 0.862
Chinese, Bejing 0.725
Japan 0.678
Chinese, South 0.607
Chinese Dai 0.428
Puerto Rican -0.007
Colombian -0.007
Gujarati Indian, Tx -0.206
Mende, Sierra Leone -0.215
Yoruba, Nigeria -0.350
US Blacks -0.402
Gambian -0.539
Bengali Bangladesh -0.594
Esan, Nigeria -0.596
Afr.Car.Barbados -0.666
Punjabi, Pakistan -0.730
Luhya, Kenya -0.921
Mexican in L.A. -0.971
Sri Lankan, UK -0.985
Indian Telegu, UK -1.039
Peruvian, Lima -1.697
If we pick the two populations for which we have accurate IQ estimates (British=100 and US Blacks=85), we can derive the other genotypic IQs. We will see that 15 IQ pts=1.61 Zs PS (Polygenic Score). Esans from Nigeria PS is 1.8 Zs lower than the British, hence 1.8/1.61*15=16.77. 100-16.77=83.23. With a similar procedure we can estimate Kenyan's IQ to be 81. But we can see that Yoruban's Nigerian's IQ is the same as US Blacks and Sierra Leone's is about 88, even higher. Conversely, Blacks from Barbados have a lower score, about 81.3.
So there appears to be differences within Africa (consistent with its status of continent with the highest and most ancient genetic variation) but also a relative disadvantage of Black from former colonies, despite their Euro admixture.
So perhaps there was some negative selection there.
Some paradoxical findings are for the Peruvian's IQ, which got the lowest estimate at 73, and Mexico at 80.
Perhaps the correction for derived alleles is too conservative and advantages Africans too much because Africans have lower baseline levels of derived alleles (which is intepreted by scholars as due entirely to drift or bias in GWAS carried out among Europeans. Some arseholes could argue that non-Africans are just generally more different from Chimps, hence having more derived (non-Chimp) alleles, hence the correction would be useless, etc). The uncorrected version does place Africans at the bottom but Euros at the top, like so. You can estimate the genotypic IQs using the same procedure I outlined above.
Total polygenic scores (uncorrected for different derived allele frequencies). Descending order. In this case I report frequencies and not Z scores.
Toscani, Italy 0.5372
Iberian, Spain 0.5338
Finland 0.53085
British, GB 0.523625
Utah Whites 0.5108
Vietnam 0.5065
Chinese, Bejing 0.5006
Chinese Dai 0.483775
Chinese, South 0.4832
Puerto Rican 0.4822
Colombian 0.4775
Japan 0.462825
Gujarati Indian, Tx 0.4339
Mexican in L.A. 0.418425
Punjabi, Pakistan 0.405875
Bengali Bangladesh 0.405825
US Blacks 0.399925
Mende, Sierra Leone 0.396475
Esan, Nigeria 0.385625
Yoruba, Nigeria 0.384
Sri Lankan, UK 0.383625
Peruvian, Lima 0.37815
Gambian 0.374875
Afr.Car.Barbados 0.37435
Indian Telegu, UK 0.37
Luhya, Kenya 0.358325
Here we have IQ= 80 for Kenya: 100-(52.3625-0.358325)/(52.3625-0.399925)*15= 80.
Gambian= 82; Peruvian= 82.4. Mexican in LA= 87.
In this instance, Hispanics score a little better.
Honestly, the consistent finding is that Africans, Amerindians and South Asians score lower than East Asians and Europeans but the precise rankings are difficult to estimate until we have a bigger sample of GWAS hits.
Thanks for the detailed replies, much appreciated.
So, where things(mainly for blacks) stand from I assume everything important is:
-Adoption studies = Inconclusive, show large(especially by age) to little effect(eg: Scarr) and studies not powerful enough.
-IQ/educational gaps African Americans = Inconclusive, NAEP closed somewhat gradually, SAT, ACT etc closed until 90s and stagnated.
-Crime/other complex traits African Americans = Inconclusive, gaps in crime closed a lot since crack epidemic in 90s but large difference still remains with males. Gaps in sexuality and other behaviours etc measured to be closing too(http://openpsych.net/forum/showthread.php?tid=211), but alas its not 100% certain the cause.
-IQ/Education of other Africans in Western countries = Inconclusive, some gaps totally erased eg: GSCE, other IQ data at age 12 largely closed and Dutch cito moderately closed... but all could be at a least a large part due to selection.
-Behaviour of other Africans in Western countries = Not enough evidence(I do not know of any long term measurements).
-Allele association = Inconclusive + challenged hereditarian additive genetic model(epigenetics: http://openpsych.net/forum/showthread.php?tid=257 ).
-Brain structure = Inconclusive.
Am I missing anything else worth a damn?
So, where things(mainly for blacks) stand from I assume everything important is:
-Adoption studies = Inconclusive, show large(especially by age) to little effect(eg: Scarr) and studies not powerful enough.
-IQ/educational gaps African Americans = Inconclusive, NAEP closed somewhat gradually, SAT, ACT etc closed until 90s and stagnated.
-Crime/other complex traits African Americans = Inconclusive, gaps in crime closed a lot since crack epidemic in 90s but large difference still remains with males. Gaps in sexuality and other behaviours etc measured to be closing too(http://openpsych.net/forum/showthread.php?tid=211), but alas its not 100% certain the cause.
-IQ/Education of other Africans in Western countries = Inconclusive, some gaps totally erased eg: GSCE, other IQ data at age 12 largely closed and Dutch cito moderately closed... but all could be at a least a large part due to selection.
-Behaviour of other Africans in Western countries = Not enough evidence(I do not know of any long term measurements).
-Allele association = Inconclusive + challenged hereditarian additive genetic model(epigenetics: http://openpsych.net/forum/showthread.php?tid=257 ).
-Brain structure = Inconclusive.
Am I missing anything else worth a damn?
I'm not so focused on African Americans in particular. That's just one ethnic group. However, they are convenient because there is a lot of data about them.
Transracial adoption studies give a pretty clear picture. E.g. http://emilkirkegaard.dk/en/?p=5663 There is a large review of studies here (http://emilkirkegaard.dk/en/?p=5672), perhaps worth doing a re-analysis of their data. I did not go into detail, but perhaps you could dig around?
The admixture evidence is pretty conclusive with regards with socioeconomic outcomes at both the regional and individual levels, and cognitive ability data at the regional level. We have been unable to locate large open datasets with individual-level data and cognitive ability, so that issue is still open. The results from the PING sample are in line with the genetic model as far as I recall.
The allele counting is 100% consistent with low African scores, but too few SNPs published so far to be sure.
The skin color sibling analysis is pretty clear too. http://humanvarieties.org/category/black-white-iq-gap/colorism/ We are in the process of redoing these analyses and writing them up for a real paper.
Transracial adoption studies give a pretty clear picture. E.g. http://emilkirkegaard.dk/en/?p=5663 There is a large review of studies here (http://emilkirkegaard.dk/en/?p=5672), perhaps worth doing a re-analysis of their data. I did not go into detail, but perhaps you could dig around?
The admixture evidence is pretty conclusive with regards with socioeconomic outcomes at both the regional and individual levels, and cognitive ability data at the regional level. We have been unable to locate large open datasets with individual-level data and cognitive ability, so that issue is still open. The results from the PING sample are in line with the genetic model as far as I recall.
The allele counting is 100% consistent with low African scores, but too few SNPs published so far to be sure.
The skin color sibling analysis is pretty clear too. http://humanvarieties.org/category/black-white-iq-gap/colorism/ We are in the process of redoing these analyses and writing them up for a real paper.
[quote][/quote]
One question about that adoption study. Are the age of adoption results fully adjusted too?
One question about that adoption study. Are the age of adoption results fully adjusted too?
Well, I hope they examined that question given that the title of the paper is "Does age at adoption and geographic origin matter? A national cohort study of cognitive test performance in adult inter-country adoptees". ;)
Conclusions. Negative pre-adoption circumstances may have persistent influences on cognitive development. The prognosis from a cognitive perspective may still be good regardless of age at adoption if the quality of care before adoption has been ‘good enough’ and the adoption selection mechanisms do not reflect an over-representation of risk factors – both requirements probably fulfilled in South Korea.
Typical blank slate conclusion. Of course, it must be the pre-adoption environment. Nevermind those 15 years or so in the Swedish school system, with Swedish health care and Swedish parents!
That being said, the test is perhaps language biased, so it is not clear how to interpret the pattern that later adoptees score lower for the non-Koreans (Table 3). It may be that the smartest children tend to be adopted first, leaving the less bright to be adopted later, if ever. Hard to say without pre-adoption cognitive data. Still, even the earliest non-Korean adoptees score 1 stanine below the last adopted Koreans (0.5 d). And there is no clear age at adoption effect for Koreans.
Of note is that the paper found zero effects of adoptive parental education and outcomes for the adoptees (Table 4), but note the strong relationship for the non-adoptees. This is the hallmark of a genetic cause.
Overall, very interesting study, but too bad they did not have more countries of origin. In fact, I asked the authors about this, but apparently they lost access to that dataset. Pity, would be able to add to the already existing transracial adoption data.
Conclusions. Negative pre-adoption circumstances may have persistent influences on cognitive development. The prognosis from a cognitive perspective may still be good regardless of age at adoption if the quality of care before adoption has been ‘good enough’ and the adoption selection mechanisms do not reflect an over-representation of risk factors – both requirements probably fulfilled in South Korea.
Typical blank slate conclusion. Of course, it must be the pre-adoption environment. Nevermind those 15 years or so in the Swedish school system, with Swedish health care and Swedish parents!
That being said, the test is perhaps language biased, so it is not clear how to interpret the pattern that later adoptees score lower for the non-Koreans (Table 3). It may be that the smartest children tend to be adopted first, leaving the less bright to be adopted later, if ever. Hard to say without pre-adoption cognitive data. Still, even the earliest non-Korean adoptees score 1 stanine below the last adopted Koreans (0.5 d). And there is no clear age at adoption effect for Koreans.
Of note is that the paper found zero effects of adoptive parental education and outcomes for the adoptees (Table 4), but note the strong relationship for the non-adoptees. This is the hallmark of a genetic cause.
Overall, very interesting study, but too bad they did not have more countries of origin. In fact, I asked the authors about this, but apparently they lost access to that dataset. Pity, would be able to add to the already existing transracial adoption data.
Typical blank slate conclusion. Of course, it must be the pre-adoption environment. Nevermind those 15 years or so in the Swedish school system, with Swedish health care and Swedish parents!
Well they have epigenetic data to back them up now(though not totally). If you recall from my epigenetic thread that environmental effects were irreversible for a few generations afterwards even under the environment to revert it. It was like that in multiple complex traits, in mammals.
Epigenetics might be able to explain away(partially or entirely) a lot of stuff, even the genetic associations I think, but I'll talk about that more specifically in the epigenetic thread, still have more research to do. Got some CRISPR editing epigenetic studies to talk about too.
You need to take all these hereditary studies and their results a notch down, it will save you trouble later on. They will use epigenetics on you sooner or later, and it will be way worse than me, because they will be insanely bias.
If you recall from my epigenetic thread that environmental effects were irreversible for a few generations afterwards even under the environment to revert it. It was like that in multiple complex traits, in mammals.
From 1840 to 1950, either semicolonized by the west or Japan, or war broken out between CPC and KMT, China's IQ score is still so high. Can you explain this with your model? It perhaps requires the "genotypic" IQ score of the Chinese to be 130 or so to be consistant with the transgenerational epigenetic repression model?
If you recall from my epigenetic thread that environmental effects were irreversible for a few generations afterwards even under the environment to revert it. It was like that in multiple complex traits, in mammals.
From 1840 to 1950, either semicolonized by the west or Japan, or war broken out between CPC and KMT, China's IQ score is still so high. Can you explain this with your model? It perhaps requires the "genotypic" IQ score of the Chinese to be 130 or so to be consistant with the transgenerational epigenetic repression model?
I'm afraid that would not work because there are many Chinese (Han) outside China. They don't have scores around 130. Of course, one could posit ad hoc that those that emigrated just happened to be -2 SD (or 1.67 if mean is 105) below the mainland Chinese. I don't know. This epigenetics hypothesis is just the latest ad hoc hypothesis to explain the gaps that has not yet been conclusively disproven. Epigenetics is an important area of science and may be useful to boost the cognitive ability of living persons (by increasing expression of boost genes via some chemical), but it will not explain much of the gaps.
@ Piffer.
Your latest polygenic results, are they corrected for derived alleles?
https://figshare.com/articles/Polygenic_selection_on_educational_attainment/3175522/1
Your latest polygenic results, are they corrected for derived alleles?
https://figshare.com/articles/Polygenic_selection_on_educational_attainment/3175522/1
@ Piffer.
Your latest polygenic results, are they corrected for derived alleles?
https://figshare.com/articles/Polygenic_selection_on_educational_attainment/3175522/1
No but I am working on it. Will upload new version soon!
This is the derived vs ancestral corrected polygenic score. I will upload the details to the Figshare manuscript and also add corrected factor scores (which seem more reliable than polygenic scores). Bear in mind that these are only educational attainment genes, I didn't include IQ genes to make the analysis more uniform.
CorrectedPS
Afr.Car.Barbados 0.4336111111
US Blacks 0.4531111111
Bengali Bangladesh 0.5047222222
Chinese Dai 0.6006666667
Utah Whites 0.4782222222
Chinese, Bejing 0.6563333333
Chinese, South 0.6336111111
Colombian 0.4766944444
Esan, Nigeria 0.4369166667
Finland 0.5366111111
British, GB 0.509
Gujarati Indian, Tx 0.4974444444
Gambian 0.44775
Iberian, Spain 0.4944444444
Indian Telegu, UK 0.4988611111
Japan 0.6424722222
Vietnam 0.61475
Luhya, Kenya 0.4418888889
Mende, Sierra Leone 0.4315277778
Mexican in L.A. 0.4805277778
Peruvian, Lima 0.4569722222
Punjabi, Pakistan 0.5128055556
Puerto Rican 0.4715833333
Sri Lankan, UK 0.4925555556
Toscani, Italy 0.48425
Yoruba, Nigeria 0.4439166667
CorrectedPS
Afr.Car.Barbados 0.4336111111
US Blacks 0.4531111111
Bengali Bangladesh 0.5047222222
Chinese Dai 0.6006666667
Utah Whites 0.4782222222
Chinese, Bejing 0.6563333333
Chinese, South 0.6336111111
Colombian 0.4766944444
Esan, Nigeria 0.4369166667
Finland 0.5366111111
British, GB 0.509
Gujarati Indian, Tx 0.4974444444
Gambian 0.44775
Iberian, Spain 0.4944444444
Indian Telegu, UK 0.4988611111
Japan 0.6424722222
Vietnam 0.61475
Luhya, Kenya 0.4418888889
Mende, Sierra Leone 0.4315277778
Mexican in L.A. 0.4805277778
Peruvian, Lima 0.4569722222
Punjabi, Pakistan 0.5128055556
Puerto Rican 0.4715833333
Sri Lankan, UK 0.4925555556
Toscani, Italy 0.48425
Yoruba, Nigeria 0.4439166667
Ok, please do upload it here too.
Edit: Woops I posted that without seeing your edit.
Edit: Woops I posted that without seeing your edit.
Thanks.
Those corrected poly scores are the 942 or just the 14?
Those corrected poly scores are the 942 or just the 14?
16 independent hits. 13 from Davies and 3 from Rietveld et al (2013). One hit was common to the two GWAS so it was not repeated, hence 13 instead of 14 for Davies. No point in using all the 942 hits really, it actually gives a biased estimate because the sites with more SNPs are given more weight, even though they represent only one signal. What's remarkable is that the corrected polygenic score is identical to the factor score obtained via factor analysis. It's as if factor analysis were clever enough to correct for it.
Please download latest version you'll find explanations: https://figshare.com/articles/Polygenic_selection_on_educational_attainment/3175522/4
Wow some of the Sub Saharan African ethnic group polygenic scores are very high in that Alfred sample for Davies 2016. Plenty of them above Europeans.
Am I reading that right?
Am I reading that right?
Wow some of the Sub Saharan African ethnic group polygenic scores are very high in that Alfred sample for Davies 2016. Plenty of them above Europeans.
Am I reading that right?
Yep, you are not reading it right. Just look at total PS or even better, factor (last column). The single PS from Davies and Rietveld have too few SNPs. The total PS has all the SNPs. Even better, look at the factor scores:
Bantu speakers -1.842
Bantu speakers -1.907
Biaka -1.402
Mandenka -1.249
Mbuti -1.637
Mozabite -0.469
San -1.886
Yoruba -1.405
Balochi -0.541
Balochi 0.144
Bedouin -0.634
Brahui -0.448
Burusho -0.266
Druze -0.394
Hazara 0.578
Kalash -0.121
Mongolian 1.068
Oroqen 0.789
Palestinian -0.418
Pashtun -0.611
Sindhi -0.533
Cambodians, Khmer 0.116
Dai 1.276
Daur 1.277
Han 1.171
Hezhe 1.262
Japanese 1.036
Koreans 1.102
Lahu 0.862
Miao 1.368
Naxi 0.102
She 1.199
Tu 1.078
Tujia 1.359
Uyghur 0.455
Xibe 1.531
Yi 0.980
Adygei -0.096
Basque -0.270
French -0.093
Italians 0.244
Italians -0.307
Orcadian 0.598
Russians -0.360
Sardinian -0.459
Maya, Yucatan -0.567
Pima, Mexico -0.406
Melanesian, Nasioi 0.027
Papuan New Guinean -0.509
Yakut 0.700
Amerindians -0.909
Karitiana -0.086
Surui -0.496
Yep, you are not reading it right. Just look at total PS or even better, factor (last column). The single PS from Davies and Rietveld have too few SNPs. The total PS has all the SNPs. Even better, look at the factor scores:
I stated I was talking about the Davies 2016 only alleles. Also some of the Sub Saharan Africans still have higher ps scores even after combination in the total and most of them close the gap substantially from Rietveld 2013.
I'm not blind.
© Copyright 2024 OpenPsych - All Rights Reserved