Hello There, Guest!  
 Previous 1 9 10 11 12 13 Next   

Ethnic Differences in the UK

(2015-Nov-23, 20:54:17)Duxide Wrote:
(2015-Nov-21, 23:38:50)Jm8 Wrote:
(2015-Nov-21, 23:28:51)Emil Wrote:
(2015-Nov-21, 22:40:34)Jm8 Wrote: The same study I linked also seems to find:

"LOW CROSS-VALIDITY OF GWAS BETAS AND POLYGENIC SCORES FOR EDUCATIONAL ATTAINMENT IN AAS(African Americans:my parenthesis) "

(title of section 6)


You realize that you are quoting my paper, right? :)


Yes.


It's unfortunate that such a statement is becoming so popular among anti-hbd folks, because in my opinion it is wrong. Differences in LD should simply reduce the observed allele frequencies differences between populations. There is no reason why these should be smaller than they appear. Correction for attenuation due to different LD should actually make the results observed by Piffer even stronger.
Edit: by observed I mean "at the sampled SNPs"..by real I meant "at the real CAUSAL variants". So correction for attenuation will increase race differences at causal variants compared to differences at the observed variants (which are in LD with the causal variants)".


It depends on which kind of error it is. If it is random error, then you are right. If it is systematic bias, then it is a problem.

However, one could investigate which direction the bias is in, I think. It should be towards the background frequency of SNPs. My guess is that is lower than the mean frequency of the cognitive ability SNPs. If so, this is downwards bias, exaggerating the results.

Can you look into it?
 Reply
(2015-Nov-24, 15:46:20)Emil Wrote:
(2015-Nov-23, 20:54:17)Duxide Wrote:
(2015-Nov-21, 23:38:50)Jm8 Wrote:
(2015-Nov-21, 23:28:51)Emil Wrote:
(2015-Nov-21, 22:40:34)Jm8 Wrote: The same study I linked also seems to find:

"LOW CROSS-VALIDITY OF GWAS BETAS AND POLYGENIC SCORES FOR EDUCATIONAL ATTAINMENT IN AAS(African Americans:my parenthesis) "

(title of section 6)


You realize that you are quoting my paper, right? :)


Yes.


It's unfortunate that such a statement is becoming so popular among anti-hbd folks, because in my opinion it is wrong. Differences in LD should simply reduce the observed allele frequencies differences between populations. There is no reason why these should be smaller than they appear. Correction for attenuation due to different LD should actually make the results observed by Piffer even stronger.
Edit: by observed I mean "at the sampled SNPs"..by real I meant "at the real CAUSAL variants". So correction for attenuation will increase race differences at causal variants compared to differences at the observed variants (which are in LD with the causal variants)".


It depends on which kind of error it is. If it is random error, then you are right. If it is systematic bias, then it is a problem.

However, one could investigate which direction the bias is in, I think. It should be towards the background frequency of SNPs. My guess is that is lower than the mean frequency of the cognitive ability SNPs. If so, this is downwards bias, exaggerating the results.

Can you look into it?


Yes, and I had already thought about this ( I would not have criticized your thought if the GWAS hits had frequencies above 50%): it depends on how it deviates from the background frequency of SNPs. As a matter of fact, the average frequency of the IQ increasing alleles is lower than that of the other alleles, which has to be 50% for obvious mathematical reasons.So the systematic error works the opposite direction from what you predicted. You should edit your paper or people will keep citing that wrong argument. Look at table 1 and table 5 (the top hits actually have a much lower than 50% freq): (https://docs.google.com/document/d/1BUmH...sp=sharing). The frequency is about 50%, slightly less so (49%). Same with top hits (table 5).
 Reply
(2015-Nov-24, 19:12:03)Duxide Wrote: Yes, and I had already thought about this ( I would not have criticized your thought if the GWAS hits had frequencies above 50%): it depends on how it deviates from the background frequency of SNPs. As a matter of fact, the average frequency of the IQ increasing alleles is lower than that of the other alleles, which has to be 50% for obvious mathematical reasons.So the systematic error works the opposite direction from what you predicted. You should edit your paper or people will keep citing that wrong argument. Look at table 1 and table 5 (the top hits actually have a much lower than 50% freq): (https://docs.google.com/document/d/1BUmH...sp=sharing). The frequency is about 50%, slightly less so (49%). Same with top hits (table 5).


I can't just edit papers to remove claims that were later found to be wrong. :P That is, I can, but I shouldn't.

Let them cite the paper. You can write a brief reply paper for Winnower presenting the numbers you mention here. That would work, no?
 Reply
(2015-Nov-24, 20:35:10)Emil Wrote:
(2015-Nov-24, 19:12:03)Duxide Wrote: Yes, and I had already thought about this ( I would not have criticized your thought if the GWAS hits had frequencies above 50%): it depends on how it deviates from the background frequency of SNPs. As a matter of fact, the average frequency of the IQ increasing alleles is lower than that of the other alleles, which has to be 50% for obvious mathematical reasons.So the systematic error works the opposite direction from what you predicted. You should edit your paper or people will keep citing that wrong argument. Look at table 1 and table 5 (the top hits actually have a much lower than 50% freq): (https://docs.google.com/document/d/1BUmH...sp=sharing). The frequency is about 50%, slightly less so (49%). Same with top hits (table 5).


I can't just edit papers to remove claims that were later found to be wrong. :P That is, I can, but I shouldn't.

Let them cite the paper. You can write a brief reply paper for Winnower presenting the numbers you mention here. That would work, no?


You could have changed it if you had not archived it already. Normally one waits to get some feedback before archiving on Winnower. Now critics will keep citing that paper, I am not sure how much it would help to have a reply paper.
 Reply
(2015-Nov-24, 20:40:14)Duxide Wrote: You could have changed it if you had not archived it already. Normally one waits to get some feedback before archiving on Winnower. Now critics will keep citing that paper, I am not sure how much it would help to have a reply paper.


I haven't archived it. "Status: Open For Review" ;)

I don't think it will be a problem. If critics cite that, fine with me. After all, I get citations! ;)

In more seriousness, it would be interesting to get to the bottom of this LD issue, that's why I'm asking you to look into it in my more detail. I don't have time.
 Reply
(2015-Nov-25, 01:49:05)Emil Wrote:
(2015-Nov-24, 20:40:14)Duxide Wrote: You could have changed it if you had not archived it already. Normally one waits to get some feedback before archiving on Winnower. Now critics will keep citing that paper, I am not sure how much it would help to have a reply paper.


I haven't archived it. "Status: Open For Review" ;)

I don't think it will be a problem. If critics cite that, fine with me. After all, I get citations! ;)

In more seriousness, it would be interesting to get to the bottom of this LD issue, that's why I'm asking you to look into it in my more detail. I don't have time.

Then just edit it. A paper under review is there to be changed according to feedback. I do not understand why you do not want to edit it...it is like posting papers on OP forum and not bothering with the reviews. It is not ethically right to post false statements just in order to get more citations, because you know such an argument is popular among anti-hbd folks.
 Reply
Quote:The study appears to claim that the alleles for those traits correlated well between Asians and Europeans but less so for Africans.

Again, it depends on the trait. And on what one means by "replicate". In Ntzani et al. table 2 (attached) ~19/24 = 80% SNPs go in same direction for Eu and Afr and ~80/97= 82% for Eu and Asians. I would have to examine in detail why Marigorta et al., seem to find different results. For them, though, "replicate" means "go in the same direction" and "meet a p-value".

Quote:Which is the African American study. Do you have a link?

The r(SNP-IQ) was ~ the same for AfrAm and EuAm. The r(SNP-edu) was lower for AfrAm but in the same direction, which is what matters. SNPg is not necessarily a good predictor of African American edu for a number of reasons so I wouldn't make much of the discrepancy.


Attached Files
.pdf   Consistency of genome-wide associations across major ancestral groups.pdf (Size: 898.94 KB / Downloads: 293)
 Reply
(2015-Nov-25, 23:45:00)Chuck Wrote:
Quote:The study appears to claim that the alleles for those traits correlated well between Asians and Europeans but less so for Africans.

Again, it depends on the trait. And on what one means by "replicate". In Ntzani et al. table 2 (attached) ~19/24 = 80% SNPs go in same direction for Eu and Afr and ~80/97= 82% for Eu and Asians. I would have to examine in detail why Marigorta et al., seem to find different results. For them, though, "replicate" means "go in the same direction" and "meet a p-value".

Quote:Which is the African American study. Do you have a link?

The r(SNP-IQ) was ~ the same for AfrAm and EuAm. The r(SNP-edu) was lower for AfrAm but in the same direction, which is what matters. SNPg is not necessarily a good predictor of African American edu for a number of reasons so I wouldn't make much of the discrepancy.


Yes, they mean 'replicate' as in obtain p<alpha and same direction.[1] This dichotomous reasoning is not good. However, they try to correct for it by correcting for power.

In my opinion this is what one wants to do if one really wants to know:

* Obtain two samples (or meta-samples) for Europeans and African Americans (or better, Africans).

* Because the African American sample is much smaller, sample random subsamples from the European samples that has the same size as the African American sample. Perform GWAS on them and the remaining sample much larger European sample. Correlate the GWAS betas from the small Euro samples with that in the large sample. Repeat the sampling for more precision.

* Calculate the mean SNP correlation for the small Euro subsamples with the larger samples. If differential LD is not a problem, then the SNP correlate for the African American sample should be the same as those from the European subsamples. Use the distribution of SNP correlations to add confidence intervals to the estimates.

The data to do this are not presently available to me, otherwise I would have done the above to find the actual cross-racial agreement.


1. It is not a good definition, but it is common. Instead they should use replicate as in obtained a result that falls within the CI of the original study. If they use 95%CI (almost always), this should happen 95% of the time IF there was no p-hacking, publication bias etc.
 Reply
Piffer seems to have reanalyzed his data and calculated somewhat higher scores for Africans in both height and iq (if my reading is right). How significant the difference is between corrected and uncorrected iq projections, I don't know.

"I computed two polygenic scores (mean population frequencies): ancestral and derived. Then I created a composite score by averaging them. This gives equal weight to ancestral and derived alleles (Piffer, 2015b).The end result is that populations with higher baseline frequencies of ancestral alleles (such as Africans) obtain a higher score after this correction, because more weight is given to ancestral alleles."

"We can see that the ranking of corrected polygenic scores for height and IQ gives higher scores to Africans compared to the uncorrected scores..."

https://topseudoscience.wordpress.com

"Derived alleles,corrected polygenic scores and height"
 Reply
For the cross-racial validity of my method, whilst I await for contrary or supporting empirical evidence, I usually appeal to Newton's four rules of reasoning, and especially rule n.2 and 3:

2. Therefore to the same natural effects we must, as far as possible, assign the same causes.
As to respiration in a man and in a beast; the descent of stones in Europe and in America; the light of our culinary fire and of the sun; the reflection of light in the earth, and in the planets.
3.The qualities of bodies, which admit neither intensification nor remission of degrees, and which are found to belong to all bodies within the reach of our experiments, are to be esteemed the universal qualities of all bodies whatsoever.
For since the qualities of bodies are only known to us by experiments, we are to hold for universal all such as universally agree with experiments; and such as are not liable to diminution can never be quite taken away. We are certainly not to relinquish the evidence of experiments for the sake of dreams and vain fictions of our own devising; nor are we to recede from the analogy of Nature, which is wont to be simple, and always consonant to itself.
 Reply
 Previous 1 9 10 11 12 13 Next   
 
 
Forum Jump:

Users browsing this thread: 3 Guest(s)