Back to [Archive] Post-review discussions

[ODP] The Elusive X-Factor: A Critique of J. M. Kaplan’s Model of Race and IQ
Admin
I see why I missed it. I looked down into the ref list until I hit a Flynn reference (the 2014 one). The 1980 one is located after some other refs, so they are not in perfect alphabetic order which I was assuming.

Here's how it looks like when I mark text to copy. The copying works fine, the trouble is only to mark the right text!
I don't refer to your work because my article is not an exhaustive review of arguments in favor of hereditarianism, and I don't think your arguments are among the most important and convincing of the various lines of evidence that exist, certainly not in the US context.


Maybe mine are not the most important or convincing arguments out there but all the other arguments are indirect. My argument is probably the only one that brings direct evidence whereas the evidence you bring is indirect and has been presented in previous papers or books, thus your paper is not very original. For these reasons I cannot approve publication.
Admin
The point of the paper is not to present all the arguments for HH, only to show that clearly HH is testable. It is a commentary paper, remember. Mentioning the Piffer method is nice, but not necessary since admixture studies work too. Sibling admixture studies are probably the strongest possible design for testing HH.
Still I don't see why it shouldn't be mentioned.
Admin
Well, there are lots of things that could be mentioned but aren't. The paper is already 40 pages. To disprove the claim that X theory is untestable, it is sufficient to present one way of testing X.
I see why I missed it. I looked down into the ref list until I hit a Flynn reference (the 2014 one). The 1980 one is located after some other refs, so they are not in perfect alphabetic order which I was assuming.

Here's how it looks like when I mark text to copy. The copying works fine, the trouble is only to mark the right text!


I'll correct the alphabetic order.

I tested the copy-paste thing on two computers, but did not encounter that problem. It's due to some software incompatibility, but I cannot do anything about it.

I don't refer to your work because my article is not an exhaustive review of arguments in favor of hereditarianism, and I don't think your arguments are among the most important and convincing of the various lines of evidence that exist, certainly not in the US context.


Maybe mine are not the most important or convincing arguments out there but all the other arguments are indirect. My argument is probably the only one that brings direct evidence whereas the evidence you bring is indirect and has been presented in previous papers or books, thus your paper is not very original. For these reasons I cannot approve publication.


I am certainly not going to add a reference to your paper if it's a condition for publication. You're making a joke out of this journal by making such a demand.

Quite purposely, my case against Kaplan's model does not rely on newfangled arguments. It's a commentary that is mostly about explaining why Kaplan's model is a non-starter due to theoretical and empirical reasons that are well-established and uncontroversial. The test of the hereditarian model that I propose is also completely uncontroversial, widely used, and definitive, unlike your method.
Admin
That is strange because I can reproduce this on two different OS (win,linux) using different PDF readers.

In any case, it is not a critical issue. Just annoying.
I am certainly not going to add a reference to your paper if it's a condition for publication. You're making a joke out of this journal by making such a demand.

Quite purposely, my case against Kaplan's model does not rely on newfangled arguments. It's a commentary that is mostly about explaining why Kaplan's model is a non-starter due to theoretical and empirical reasons that are well-established and uncontroversial.


I only requested you to cite my work because it explicitly addresses one of the issues raised by Kaplan, which your paper fails to consider: This is what he wrote: "But positing, without any evidence, systematic differences in hundreds or thousands of genes of small effects is surely no more plausible than positing multiple environmental differences with small effects! as noted above, we do know of some environmental differences between the populations that are verifiably associated with differences in performance on IQ tests and related measures; we know of no genes that are so-associated."

In order to counter Kaplan's argument, in my first post I had asked you to cite Ward et al's study which replicated Rietvald et al's 3 hits. This you didn't do. I had also asked you to cite my work, which again counter Kaplan's argument. This you didn't do.

If I am making a joke out of this journal it's simply because the review is open. I've met with so many unreasonable requests by reviewers but none could see them! You should be thankful that this journal is based on an open review system. Besides, I am not the journal. I think you're making a joke out of your paper by stubbornly refusing to address the most important argument in Kaplan's article. He's gonna laugh out loud.
I only requested you to cite my work because it explicitly addresses one of the issues raised by Kaplan, which your paper fails to consider: This is what he wrote: "But positing, without any evidence, systematic differences in hundreds or thousands of genes of small effects is surely no more plausible than positing multiple environmental differences with small effects! as noted above, we do know of some environmental differences between the populations that are verifiably associated with differences in performance on IQ tests and related measures; we know of no genes that are so-associated."

In order to counter Kaplan's argument, in my first post I had asked you to cite Ward et al's study which replicated Rietvald et al's 3 hits. This you didn't do. I had also asked you to cite my work, which again counter Kaplan's argument. This you didn't do.

If I am making a joke out of this journal it's simply because the review is open. I've met with so many unreasonable requests by reviewers but none could see them! You should be thankful that this journal is based on an open review system. Besides, I am not the journal. I think you're making a joke out of your paper by stubbornly refusing to address the most important argument in Kaplan's article. He's gonna laugh out loud.


As I wrote earlier, Kaplan could not have cited Ward et al., because that study was published after his paper. I can hardly fault Kaplan for his lack of clairvoyant abilities. Moreover, the three confirmed hits explain very very little of total variance, so they cannot be used to say anything about group differences -- unless you make all sorts of strong assumptions that may or may not be correct.

In my paper, I only use arguments and methods that I thoroughly understand myself, or at least ones that are widely accepted by relevant experts. Your method seems ok to me, but I readily admit that my understanding of population genetics is limited, so I cannot vouch for it. The fact that your method has not been reviewed and adopted by experts in the field means that I cannot accept it on authority, either.
You don't have to fault Kaplan for his lack of clarvoyant abilities. But it'd be good if you could show that his beliefs are outdated. You could state that although at the time of Kaplan's writing his statement that "we know of no genes that are associated with IQ or related measures" was true, today it's wrong. You don't have to cite my papers if you think they're not mainstream enough but at least cite Ward et al and Rietvald et al. After that I'll approve.
The genetic correlations indicate that the same genes explain most of the heritability of seemingly unrelated different abilities, e.g. verbal and perceptual ability. This is consistent with the g model. It does not prove the validity of the model, just increases its plausibility. Modelling g explicitly would not prove the g model, either, just potentially increase its plausibility even more.


Normally, a better model fit increases the likelihood of this model, and it's why it is selected against others. Of course, as you say, CFA modeling is not aimed to "confirm" a model. That's because a theory cannot be proven in science. We can only reject theories, not confirm them; but to the extent we continuously fail to disconfirm a given theory, it is becoming more and more plausible. But it's still not a proof in itself. So, any methods aimed at "hypothesis-testing" approach should only disprove models (or theories underlying these models). This can be done when our model has a worse fit than the alternative model. In the case of g vs non-g, it's not clear at all what to conclude. At the very least, because g model supports the weak version of Spearman, I accept the idea that g model is superior than non-g model, but without superior fit for g model, I cannot conclude the evidence is strong. But that it is only weak, or meager proof in favor of g.

You will never have a single test that will determine what the correct model is. The are always alternative models in CFA that fit equally well. You will have to look at the big picture, all the evidence.


Alternative models, if I'm not mistaken, have to do with models that are mathematically equivalent but conceptually different, e.g., reversing path arrows. When the df of the models is the same, you will expect equal fit. See below.

https://groups.google.com/forum/#!msg/lavaan/KfInb51WA2Y/eezpGsl_YesJ
This is one of the quirks of CFA/SEM. In theory, this second-order model should provide exactly the same fit as the (correlated) three-factor model, since the number of free parameters is the same. But it often fails. Sometimes, adding std.lv=TRUE may help, but not always.


By equal fit, I say "exactly" the same. In Dolan (2000) the models are not the same, and the df weren't the same either. The model fits however are very similar. But not "exactly" equal. But this is sufficient to conclude there is no proof in favor or disfavor of g model.

With regard to the other methods, I agree that it is more in accords with g models than not. Unfortunately, Dolan and others believe these (e.g., MCV, PCA) are weak methods, and they think we should give more weight on MG-CFA.

My point of citing Wicherts on the Flynn vs. b-w gap is that the causal processes behind these gaps are different, as indicated by MI analyses.


Yes, except that his analysis is false. It can't show you what he wanted to show. Again, I could have agreed if loading invariance is violated, which it isn't. It's only the intercept, and the direction of bias is both-sided, tend to cancel out. Imagine for example that the BW difference is somewhat biased (with cancel out at the total test score) but it is modest (as assessed by modest model fit decrement). Now, you took another test for these same groups, and find strong violation of MI at intercept level, but the mean score difference is the same. Why ? The answer is because the biases are stronger. Instead of -1 IQ points (against blacks) for subtests 1-4 and -1 IQ points (against whites) for subtests 5-8, now you have -5 IQ points for subtests 1-4 and -5 points for subtests 5-8. Given this the total score difference is the same. You cannot say, as Wicherts claimed, that the IQ difference is biased because there are larger IQ losses for either groups in some or all subtests. To be sure, it's nonsense to speak about "bias" without mentioning the direction of bias, when there is one such pattern.

The purpose of his analysis was that BW can't be explained by psychometric bias, and that Flynn gain can be explained by psychometric bias, thus his conclusion that they both are unrelated. And yet, it's untrue that FE gains can be explained by psychometric bias. If so, the FE gains would have vanished. I don't see it.


---
---

EDIT

After reading Piffer's comment, I think I agree with him. It's not a question of how much variance they explain (assuming R² is an effect size, that is...) but just to show that we know now of such genes.
In my paper, I only use arguments and methods that I thoroughly understand myself, or at least ones that are widely accepted by relevant experts. Your method seems ok to me, but I readily admit that my understanding of population genetics is limited, so I cannot vouch for it. The fact that your method has not been reviewed and adopted by experts in the field means that I cannot accept it on authority, either.


I agree with Dalliard. I don't think that Davide's paper ought to be cited; the results are weak and the method hasn't been vetted. Also, citing Ward et al and Rietvald et al would require an extended discussion which wouldn't well fit with the remainder of the article. Davide might instead pen his own critique of Kaplan's paper.

I don't understand Meng Hu's criticism. Regarding MCV and MI, Dalliard's logic makes sense. (I have made the same arguments myself.) In regards to MCV:

1. Environmental effects such as schooling tend to be most pronounced on the least g-loaded sub-tests.
2. The B/W gap shows the reverse pattern.
Ergo: The B/W gap is not due to these types of effects.

In regards to MI:

1. When MI, group differences are psychometrically unbiased.
2. Effects such as Stereotype and the Flynn Effect induce psychometric bias.
3. The B/W gap generally has been found to be psychometrically unbiased i.e., MI has been found to hold.
Ergo: The B/W gap is not due to effects such as the Flynn and Stereotype Effect -- which seemingly are x-factor ones.

Of course the Flynn effect can partially be in g and race gaps, such as international ones, that show psychometric bias can be likewise. But that's not the point. What is is that if the race gaps in the U.S. are due to measurement non-invariance or anti-Jensen Effect inducing variables then the differences should consistently show measurement non-invariance and anti-Jensen Effects. But they don't.

Basically, I think that MH misunderstood the argument. Also, I strongly disagree with his characterization of the evidential status of SH. I consider SH to be well supported. I outlined 14 lines of evidence here. Also, I explained that different models in fact can be tested with MCV here.

What emerges robustly from the above analysis is that g-loadings mediate the association between cultural-loadings and the magnitudes of the Black-White gaps. I have argued elsewhere that such mediation more strongly evidences g-differences than do bivariate correlations between g-loadings and the magnitude of the subtest differences. Mediation literally puts g at the center of things and makes less plausible arguments that the correlation between group differences and g-loadings are driven by e.g., broad factor differences. That is, they make less plausible the specificity critique of the method of correlated vectors.

It might be helpful to illustrate the above criticism of MCV. Below shows the correlation between the WISC IV deaf /hearing differences and g-loadings. The deaf/hearing differences were taken from Krouse (2012), who found that the D/H differences were measure non-invariant. As seen, this correlation was positive and significant. On further analysis, though, this correlation turned out to be driven by verbal factor differences. The D/H difference was largest on the verbal factor and the verbal factor happened to have a higher g-loading than the non verbal factors. This resulted in a Jensen Effect, which disappeared when one regressed out the effect of factors.

That the Jensen Effect was driven by verbal factor differences was confirmed by an analysis of 9 studies on the H/D performance subtest differences. The values were taken from Braden (1990). All of these analyses showed moderate to strong negative correlations between the magnitude of the group differences and performance test g-loadings. In comparison, for all 6 WISC samples discussed below, the r (g x BW) was highly positive for both performance and verbal subtests. In fact, I was unable to find any mediating non-g factor.


The extended MCV logic is:

1. g-driven differences show Jensen Effects.
2. Alternatively, a Jensen Effect could be purely spurious. But meta-analysis rules this implausible.
3. Alternatively, a Jensen Effect could be driven by a gap in g-loaded broad or narrow factors. But multivariate MCV rules this improbable.
4. The B/W gap shows Jensen Effects.
Ergo: SH holds.

To put this another way, if all broad factors or all combinations of narrow factors show a Jensen Effect, then it is highly unlikely that the group difference is not largely due to g when the group difference shows a Jensen Effect. This specificity critique is more magical thinking; I would like its proponents to specify which non-g factors are driving the r (B/W x g).
...

I approve publication. Who doesn't and for what precise reason? Let's get this show on the road.
I agree with Dalliard. I don't think that Davide's paper ought to be cited; the results are weak and the method hasn't been vetted. Also, citing Ward et al and Rietvald et al would require an extended discussion which wouldn't well fit with the remainder of the article. Davide might instead pen his own critique of Kaplan's paper.

I approve publication. Who doesn't and for what precise reason? Let's get this show on the road.


Chuck, your criticism seems out of place. I do not know why you claim my results are weak, this comment sounds unnecessarily hostile especially because it's not backed up by sound reasons. I do not think my results are weak for many reasons which I am not gonna outline here, but my method has shown extraordinary predictive power both within and across continents (effect sizes are very high, with correlations to phenotypic IQ in the 0.9-0.95 range), after controlling for migrations and socioeconomic variables. The only weakness is that it's based on only 4 SNPs. However, I am not gonna turn this thread into an apology of my work.
Apart from this, I explained that the author should cite Rietveld et al. + Ward et al study to show that Kaplan is wrong in claiming that there are no genes with replicated effects on intelligence. This is important because: this claim is very common and is not only present in Kaplan's paper, but is made by a wide range of anti-hereditarians. 2) The strength of Kaplan's argument relies on this claim too.
Since this is a commentary on Kaplan's paper, it'd be lame to ignore this very important point. I do not think that citing these works would require an extensive discussion. It'd just be enough to state that these GWAS hits have been replicated.
Admin
I agree that citing Rietvald et al + Wald et al is in order. No need for a long discussion, just that previous studies were too small giving a rate false positive rate.
Chuck, your criticism seems out of place. I do not know why you claim my results are weak, this comment sounds unnecessarily hostile especially because it's not backed up by sound reasons.


Harpending correctly noted:

Actually, there is some evidence in on that question. The big education GWAS study last year found three replicable SNPs that influence educational performance. Here is their distribution:

SNP Effective Allele Sign Yoruba Japan China Europe
rs9320913 A + 0.181 0.326 0.419 0.508
rs11584700 A – 0.958 0.633 0.700 0.758
rs4851266 T + 0.042 0.455 0.578 0.358

Three could be coincidence. Drift or linkage could pull the frequency of a given SNP in a different direction from the general trend caused by selection. If you had 17 out of 20, or 40 out of 50, you’d have a bulletproof case.


The results are weak because, as yet, you only have 3 to 4 well replicated alleles. I am certain that most experts in the relevant fields would agree. Would you like to bet?

Apart from this, I explained that the author should cite Rietveld et al. + Ward et al study to show that Kaplan is wrong in claiming that there are no genes with replicated effects on intelligence.


Ward et al. replicated educational alleles discussed by Rietveld et al. Yet Kaplan discusses the etiology of IQ differences. The difference might seem nitpicking to you, but I can assure that Kaplan and Co. will see things differently. For this reason in addition to that previously given, I concur with Dalliard's opinion.
Harpending correctly noted:

Actually, there is some evidence in on that question. The big education GWAS study last year found three replicable SNPs that influence educational performance. Here is their distribution:

SNP Effective Allele Sign Yoruba Japan China Europe
rs9320913 A + 0.181 0.326 0.419 0.508
rs11584700 A – 0.958 0.633 0.700 0.758
rs4851266 T + 0.042 0.455 0.578 0.358

Three could be coincidence. Drift or linkage could pull the frequency of a given SNP in a different direction from the general trend caused by selection. If you had 17 out of 20, or 40 out of 50, you’d have a bulletproof case.


The results are weak because, as yet, you only have 3 to 4 well replicated alleles. I am certain that most experts in the relevant fields would agree.

Apart from this, I explained that the author should cite Rietveld et al. + Ward et al study to show that Kaplan is wrong in claiming that there are no genes with replicated effects on intelligence.


Ward et al. replicated educational alleles discussed by Rietveld et al. Yet discussed is the etiology of IQ differences. The difference might seem nitpicking to you, but I can assure that Kaplan and Co. will see things differently.


It looks like you've not read my papers because you are reporting a comment by Harpending, which reports allele frequencies only for 4 populations. Instead, I reported these frequencies for 14 populations from 1K Genomes and 50 from ALFRED and showed that they produce a factor accounting for large part of the variance. Where have you been all this time? Linkage is no explanation, as these 3 SNPs are located on different chromosomes and are unlinked. Drift is also unlikely because it does not push frequencies of 4 alleles with the same phenotypic effect in the same direction. Drift does not have a consciousness and does not know which phenotypic effect an allele has...it'd be more likely as an explanation if I had selected from the 4 SNPs, 4 alleles randomly out of the total 8, instead of those which increase the trait in question. I've also controlled for migratory effects and other possible statistical artifacts, plus socioeconomic variables.
For an update, look here:
Estimating strength of polygenic selection with principal components analysis of spatial genetic variation
Davide Piffer
doi: http://dx.doi.org/10.1101/008011

Again, I do not wanna risk turning this thread into an apology of my work. I already said that I will approve this paper even if my work is not cited so your comments reveal not only lack of familiarity with my work but they're also out of context (revealing lack of comprehension of my previous posts)
It's also not true that Kaplan et al only discussed IQ differences.
He states "as noted above, we do know of some environmental differences between the populations that are verifiably associated with differences in performance on IQ tests and related measures; we know of no genes that are so-associated." Ward et al.'s measured performance in math and reading, which are IQ "related measures"
Rietveld et al. showed that these 3 alleles are significantly associated with g (the effect size being bigger than for educational attainment) in a Swedish sample.
I already said that I will approve this paper even if my work is not cited so your comments reveal not only lack of familiarity with my work but they're also out of context (revealing lack of comprehension of my previous posts)


I don't believe that I suggested otherwise. I simply said: "I agree with Dalliard. I don't think that Davide's paper ought to be cited; the results are weak and the method hasn't been vetted. " I didn't say e.g., "Davide's most recent demands are unjustified". You're reading too much into this.

As for the other point, you agree that the method hasn't been vetted, so you only disagree with my statement that the results are weak. I said that they were because the number of alleles was few -- thus the evidence for (polygentic) selection in the said directions is still tentative. What was the joint probability for the six correlations (4 alleles, n = 14)? As it is, I asked several population geneticists about the results in relation to the letter claiming that "We are in full agreement that there is no support from the field of population genetics for Wade’s conjectures,” and they said the same: weak results and unconfirmed method. When I get a chance, I will email the group of signatories and let you know their response.

Whatever, we agree that this need not be cited.

It's also not true that Kaplan et al only discussed IQ differences.
He states "as noted above, we do know of some environmental differences between the populations that are verifiably associated with differences in performance on IQ tests and related measures; we know of no genes that are so-associated." Ward et al.'s measured performance in math and reading, which are IQ "related measures"Rietveld et al. showed that these 3 alleles are significantly associated with g (the effect size being bigger than for educational attainment) in a Swedish sample.


Kaplan said:

Despite large-scale and intensive searchers, very few alleles associated with differences in IQ test performance (within ‘‘White’’ populations) have been identified. While the contemporary estimates for the heritability of IQ test performance cited by hereditarian researchers range in the .5–.8 zone,21 searches have uncovered genes associated with no more than a tiny fraction of the variance (see e.g. Plomin 2013; Plomin et al. 2013), and some researchers claim that most of these are likely false-positives (see e.g. Chabris et al. 2012)....

But positing, without any evidence, systematic differences in
hundreds or thousands of genes of small effects is surely no more plausible than positing multiple environmental differences with small effects! Indeed, as noted above, we do know of some environmental differences between the populations that are verifiably associated with differences in performance on IQ tests and related measures; we know of no genes that are so-associated.


And your point is that Ward et al. and Rietvald et al.'s results establish that some specific genes are "verifiably associated with differences in performance on IQ tests and related measures".

1. I'm not sure that Ward et al. and Rietvald et al.'s establishes this. It depends on how you read AND (and "verifiable"). You're reading AND as OR. I see it as a conditional statement: "verifiably" associated "IQ" alleles needed.

2. The point about no verifiably associated IQ alleles is not significant to Kaplan's argument; it's more of a passing swipe.
Actually it is false that positing systematic differences in hundreds or thousands of genes of small effects is no more plausible than positing multiple environmental differences with small effects.
Polygenic selection predicts that when a phenotype is selected for, the frequencies of hundreds or thousands of alleles across the genome will be affected as well. So a single cause can affect many genes, so insofar as they have the same phenotypic effect, the genes are not independent of one another and they cannot be treated as such. A conditional probability approach is necessary. But whatever, Daillard has admitted he does not know population genetics well enough so he's not qualified to comment on my work (which makes his former remark that my work is not very important all the more meaningless, given that he's not qualified to issue such a judgment).
I guess Chuck and I disagree on the the interpretation of the AND, whether it should be interpreted as the logical operator "and" (his interpretation) or the more colloquial version (or). Even if we disagree on this, I don't understand where all this stubborness is coming from. It certainly wouldn't hurt to add a paragraph reporting Ward et al.'s + Rietvald et al's results, given that they report genes associated with measures related to IQ and in one instance (Rietveld et al) also directly related to g (at least in a subsample). This is not of major importance but the refusal to do so seems totally unjustified to me.
Regarding your judgement of my method, of course it's unconfirmed, because it's new. All methods, from the most brilliant to the dodgiest, had to be unconfirmed at some point. This does not mean it won't be confirmed in the future.
Admin
As for approvals:
Emil - Approve
Chuck - Approve

2 out of 4.

The forum needs some software that can display the current number of approvals. However, since we are possibly the first journal to use a forum for this purpose, we will have to write the code ourselves.
When I get a chance, I will email the group of signatories and let you know their response.


What are you gonna email them? My latest paper I hope (assuming they read it carefully) http://dx.doi.org/10.1101/008011

Whatever their feedback I won't consider it as the final word on the subject.