“Fascinating. IQ is clearly quantitative. So what definition of trait are you using to attempt to justify your assertion?”

How is it quantitative? IQ tests test learned skills and knowledge, not ‘intelligence’. IQ tests aren’t construct valid so you can’t logically state that they test ‘intelligence’.

“Do you think if you repeat that often enough it will magically become true? Are you saying the brain (rather biological, I would say) has NOTHING to do with IQ? Do you honestly not understand the strength of words like “no” in your usage?”

Yes I understand, I wrote it. Correlations arise because the brain is part of the body taking the IQ test. How do you show the correlations are causal? They’re not.

“Actually (and leaving aside the obvious false dichotomy) the logic does not work that way. If the EEA is false then the twin study method does not give valid results. That does not mean “MZ and DZ trait differences comes down to environment.””

How many truth values are there? Can the EEA be ‘kinda true, half false half true’?

“Even Ken Richardson understands that”

Similarity of environments cannot be directly measured for myriad reasons, including but not limit to: not knowing trait-relevant environments and similar treatment of MZ twins.

And I’m not Ken Richardson. I’ll go a step beyond him and say it’s false. Also check the paper I cited above by Joseph et al and see the paper on schizophrenia and the EEA which Richardson coauthored.

“Look, you are obviously not engaging in this topic in a sincere and intellectually rigorous fashion.”

Why do you say this? Because I said there’s no biological basis for IQ or that IQ isn’t a quantitative trait like height? That assumption is because heritability increases as age does. But heritability only increases with age due to how the tests are constructed. Change the test, change these outcomes, too. Test construction is a big thing regarding IQ, as well as the non-construct validity of this so-called ‘construct’.

Let’s say I had a test for ESP. The test is better than chance. But it’s not validated—we’re not sure that the test tests what is purported to test. Does my ESP test test ESP? Now think the same for IQ. IQ tests aren’t construct valid. Can you logically state that it tests ‘intelligence’ and not learned skills and knowledge?

You’re assuming that IQ is a quantitative trait like height but this assertion is not valid.

Fascinating. IQ is clearly quantitative. So what definition of trait are you using to attempt to justify your assertion?

There is no biological basis to ‘IQ’.

Do you think if you repeat that often enough it will magically become true? Are you saying the brain (rather biological, I would say) has NOTHING to do with IQ? Do you honestly not understand the strength of words like “no” in your usage?

It’s simple: if the EEA is true then differences between MZ and DZ trait differences come down to genetics. But if the EEA is false then MZ and DZ trait differences comes down to environment.

Actually (and leaving aside the obvious false dichotomy) the logic does not work that way. If the EEA is false then the twin study method does not give valid results. That does not mean “MZ and DZ trait differences comes down to environment.”

Even Ken Richardson understands that: https://onlinelibrary.wiley.com/doi/pdf/10.1348/000709904X24690
Emphasis mine. Note the impressive string of weasel words (and the contrast to your statement that I quoted).

Here, we show that the most crucial of these, namely, the equal environments assumption (EEA), may not hold. Consequently, differences in twin correlations might be at least partly explained by treatment effects from parents, teachers, peers, and so on.

Look, you are obviously not engaging in this topic in a sincere and intellectually rigorous fashion. I don’t know why that is, but that does not really matter. Unless you have something new to bring to this conversation, I think I have said enough.

You’re assuming that IQ is a quantitative trait like height but this assertion is not valid. There is no biological basis to ‘IQ’.

The EEA is false. Therefore so-called heritability is shared environment. No argument you make or attempt to make will make the EEA true. Twin studies are useless as are heritability estimates.

It’s simple: if the EEA is true then differences between MZ and DZ trait differences come down to genetics. But if the EEA is false then MZ and DZ trait differences comes down to environment. The EEA is false (http://logosjournal.com/2015/joseph-twin-research/) therefore heritability equals shared environment and you cannot logically make genetic inferences from the data.

IQ isn’t a real biological reality. Height is.

And what does that have to do with the EEA?

This is the part where I wonder if you honestly believe the arguments you are making. If so, it does not speak well of you.

“Please elaborate on how “environmental influences are shared to the same extent by monozygotic and dizygotic twins” applies for height but not for IQ.”

IQ isn’t a real biological reality. Height is.

“Modeling reality is all about what assumptions are “good enough.””

Well these assumptions aren’t ‘good enough’. If the EEA is false then differences between MZ and DZ twins comes down to environment.

“fetishization of the EEA”

How is it “fetishization’? It’s either true or false. If it’s false differences between MZ and DZ twins is environmental.

“essentially all models are wrong but some are useful.”

The CTM isn’t useful because the EEA is false. Therefore heritability is shared environment and we cannot logically state that genes cause differences between MZ and DZ twins.

there are no EEA arguments for height

Really?! Please elaborate on how “environmental influences are shared to the same extent by monozygotic and dizygotic twins” applies for height but not for IQ. While keeping in mind your all or nothing statement:

No false dichotomy exists. The differences between MZ and DZ twins either comes down to genetics (EEA is true) or environment (EEA is false); it can’t be ‘half true/false’ or ‘one-quarter true/ three-quarters false’. It’s either the assumption is false or it isn’t, and the assumption is false.

You really aren’t very analytical, are you? Modeling reality is all about what assumptions are “good enough.” The discussion surrounding this quote is useful: https://stats.stackexchange.com/questions/57407/what-is-the-meaning-of-all-models-are-wrong-but-some-are-useful

It is fascinating (though sometimes also mystifying) to me to see what some people consider compelling arguments.

BTW, your fetishization of the EEA in this context is a great example of an isolated demand for rigor: http://slatestarcodex.com/2014/08/14/beware-isolated-demands-for-rigor/

You should email that to him. He responds to emails.

Doesn’t the h2 estimate assume that g doesn’t interact with e? Like, they work additive independent of each other?

“You really don’t see how that is a false dichotomy and the differences between MZ and DZ twins can be on a continuum from 0-100% on the environmental and genetic influence split?!”

No false dichotomy exists. The differences between MZ and DZ twins either comes down to genetics (EEA is true) or environment (EEA is false); it can’t be ‘half true/false’ or ‘one-quarter true/ three-quarters false’. It’s either the assumption is false or it isn’t, and the assumption is false.

http://logosjournal.com/2015/joseph-twin-research/

“I don’t think it is worth taking Ken Richardson’s work seriously until you explain why his ideas did not cause problems with the Compressed Sensing results for predicting out of sample height from the genomes.”

Height is not IQ and there are no EEA arguments for height, as utu said. Height and IQ are two different things, despite claims to the contrary by people like Hsu who liken them to be similar.

Arguments have premises and conclusions. Hence, no argument was made.

There is no point. The brain is a part of the person taking the test, hence the correlations. none have shown to be causal regarding physiological process —-> IQ score.

Go on.

“Argument over.”

How?

Thank you! Going through that in detail will take more time than I have this instant, but some quick thoughts.

Page 4 has this description (equation rewritten due to special characters):

Cohorts were asked to estimate this regression equation for each measured SNP (we drop the SNP subscript j here to avoid notational clutter):
(1) EduYears = Beta0 + Beta1 * SNP + PC * Gamma + B * Alpha + X * Theta + epsilon,
where SNP is the allele dose of the SNP; PC is a vector of the first ten principal components of the variance-covariance matrix of the genotypic data, estimated after the removal of genetic outliers; B is a vector of standardized controls, including a third-order polynomial in age, an indicator for being female, and their interactions; and X is a vector of study-specific controls. Specifically, in X, study analysts were encouraged to include dummy variables for major events such as wars or policy changes that may have affected access to education in their specific sample. Mixed-sex cohorts were additionally asked to upload separate regression results for men and women.

It would be interesting to see how the PCs they observed corresponded to those seen in studies intentionally looking at variation between populations.

Am I wrong to infer that this equation could be used to estimate coefficients for the variables being controlled for? Though since the study is not designed for those tests it is unclear how reliable those coefficients would be (the coefficient SDs should give a good clue, right?).

Page 21 has some discussion of polygenic scores with and without PC adjustment.

There is more detail on polygenic prediction starting on page 107. There they reference Supplementary Table 5.2 for results.
The paper itself is at https://www.nature.com/articles/nature17671 along with many supplementary tables in Excel.
Looking at that table I don’t see any coefficients given for the control variables. The low R^2 values seen leads me to conclude they did not include the control variables in the R^2. Does anyone know for sure if this is the case? If so, it would be very interesting to know what the R^2 was for the baseline model.

Page 109 has a comparison with Rietveld et al. (2013). Does anyone know if they used the same PC control methodology?

Okbay et al. reference https://www.ncbi.nlm.nih.gov/pubmed/16862161 which I should probably check out for more detail if needed.

P.S. The detailed contribution breakdown on pp. 125-133 (!) was something.

Peter,

Thanks for your comment. You say that ” researchers use such extreme corrections to prevent this statistical bias from affecting their results” and that “GWAS typically removes the first fifteen principal components of the SNP (single nucleotide polymorphisms) correlation matrix,” but did the authors of this particular study take such precaution against statistical bias? The description of methods does not, so far as I can see, say that they did.

As to the general conclusion, namely, that:

Is it not the case that the claim:

“… genetic …. differences between school types are primarily due to the heritable characteristics involved in pupil admission” merely a tautology?

As to the conclusion that independent schools offer no value added, it reminds me of the categorization of private schools by Mr. Levy of the Gargoyle Employment Agency in Evelyn Waugh’s novel Decline and Fall:

We class schools, you see, into four grades: Leading School, First-rate School, Good School, and School.

Having myself, in the days of Evelyn Waugh’s prime, attended many different schools, including one “Leading School” one public grammar school, and a couple of mere “Schools,” it seems doubtful to me that this study, which treated all fee-paying schools as a group, could have anything useful to say about the potential educational added value of a private school education.

Including the principal components in the regression is statistically more efficient than prewhitening before running the regression. If interested I suggest you look at the Supplementary Information appendix in Okbay et alia’s important article in Nature (2016):

https://media.nature.com/original/nature-assets/nature/journal/v533/n7604/extref/nature17671-s1.pdf

GWAS typically removes the first fifteen principal components of the SNP (single nucleotide polymorphisms) correlation matrix.

This is new to me. Can you give a reference describing how this is done exactly?

In a quick search the closest thing I found was from: https://www.nature.com/articles/mp2017153

GWAS was performed using PLINK v1.9 (ref. 29) testing for associations between SNPs and alcohol consumption in unrelated individuals with location of UKB assessment centre, genotyping batch and 15 principal components included as covariates

That sounds different from your statement and actually sounds like a quite direct way to look at group differences if the first PCs correspond to racial groups as they typically do in studies of diverse populations.

Perhaps I am reading it wrong though and they mean the covariates are corrected for rather than being considered as explanatory variables. If so, it seems to me they would be removing signal more than noise by controlling for the PCs. Though it would help correct for population differences.

Elsewhere in the paper we see:

Individuals were removed from the present study based on non-British ancestry (within those who self-identified as being British, principal component analysis was used to remove outliers, n=32 484)

which seems more in line with your point.

And also:

Four principal components were fit as fixed-effect covariates to control for population stratification.

GWAS typically removes the first fifteen principal components of the SNP (single nucleotide polymorphisms) correlation matrix. This removes any strong ethnic common features in the SNP’s since they are in the first few principal components. Many studies also rely on the European-only subsample as an additional control. Your comment is theoretically correct, which is why researchers use such extreme corrections to prevent this statistical bias from affecting their results.

Given the multi-racial character of Britain’ school-age population, is it not possible — probable even, that the relationship between genome-wide polygenic scores (GPS) and scholastic achievement/aptitude reflect differences in GPS among the highly culturally distinct racial groups comprising Britain’s often poorly integrated immigrant communities? In that case, the GPS may have little or nothing to do with academic aptitude.

Apparently, no attempt was made to control for racial differences, which one might expect to differentiate among students as well as the genetic markers. To prove otherwise, one would need to show that within, as well as among, racial/cultural groups, GPS reflects scholastic achievement/aptitude.

A new paper on educational attainment and genetic differences across private/state school student cohorts in Britain:

https://www.nature.com/articles/s41539-018-0019-8

Yes, study of the anatomical and physiological correlates of mental activity is surely the way forward.

I was unable to get access to the full text of the Haier and Jung paper, but there’s no end of related stuff of great interest.

For example this paper indicating a relationship between cerebral blood flow and creativity, which is consistent with Newton’s account of his discoveries which entailed “keeping the question constantly before the mind until, little by little, it opened up.”

This effect, which should perhaps be called the “Newton creativity factor,” may explain the apparent independence of IQ and creativity. Indeed, have you not noticed how some people of the highest academic aptitude, and therefore presumably of high IQ, can seem stumped when it comes to entertaining a hypothesis, indicating that the Newtonian thought process is quite alien to them, whereas to someone such as Richard Feynman, a man of reputedly modest IQ, a brainfull of hypotheses was the normal state of mind.

And here it is asserted that artistic and scientific creativity are attributable to activity in different brain regions, which comes as no surprise to the scientific brain baffled by the phenomenon of modern “art.”

Such studies reveal great potential for the quantification of mental capacity, and the observation of mental development and the impacts on mental development of culture, education, nutrition, etc.

and it is only kids from near the top of the pile who take the College Board SAT tests

That depends on how narrow/broad your definition of “near the top” is. Based on the NYT graphic below about 1.67 million students took the SAT in 2012. A similar number took the ACT. Eyeballing Figure 2. in https://www.census.gov/prod/cen2010/briefs/c2010br-03.pdf we can estimate about 4.3 million American 15 year olds in 2010.

Those numbers give a range for those taking either the SAT or ACT of 39%-78%. I believe this does ignore foreign students taking the tests. I am not sure how multiple attempts are accounted for.

I don’t know how those numbers accord with your idea of “near the top.”

From another perspective, this statistic surprised me: https://edsource.org/2014/evaluation-shows-number-of-students-not-taking-sat-act-or-ap/63320

No students are taking SAT, ACT or Advanced Placement exams in 14 percent of California’s high schools

Related: https://www.npr.org/sections/ed/2015/10/07/446322589/surprisingly-few-urban-high-school-students-take-the-act-or-sat

For teenagers headed to a four-year college, taking a standardized entrance exam such as the ACT and SAT is typically a requirement. But it’s far from a universal experience.

In 50 of the largest U.S. cities, examined in a new report from the University of Washington, Bothell’s nonpartisan Center on Reinventing Public Education, fewer than 1 in 3 students takes either of those tests in a given year.

The rate of taking the SAT or ACT in those cities topped out in Memphis, at just 30 percent. In three-fifths of the cities, it was less than half that.

SAT and ACT test takers by year:

There are some subgroups of the tests which have higher correlations. For example, look at the correlations for #13 Oral Vocabulary with tests 20-28.

Yes, they include “Oral vocabulary,” “listening comprehension,” and “verbal analogies.” Almost certainly a lot of overlap there in terms of the underlying neurological processing. What that suggests is that the tests are not well designed to correspond with the function of particular neurological structures.

This is a problem, even with such seemingly distinct areas as math and verbal, since math questions and concepts generally have a verbal component. Has anyone, I wonder, done brain imaging of students taking IQ tests. It could be revealing, not that I recommend using kids as guinea pigs. The value of such information would be to design better tests, i.e., tests that measure the performance of distinct, as opposed to overlapping, brain areas.

And, as noted earlier, assortative mating must account, in part, for correlation between SAT math and SAT verbal, especially at the top end of the socioeconomic pile (and it is only kids from near the top of the pile who take the College Board SAT tests). In other words, smart people marry other smart people, attracted by the money if not the brains, and you cannot be really smart unless you can use both words and numbers with skill. My prediction, therefore, is that the math:verbal correlation will decline from the top down, to the midway point. Beyond that, with deepening dimness, math and verbal scores might be expected to converge again.

See, recognition that intelligence is not a unitary phenomenon provides a whole new way of looking at psychometrics. The goal should be not convergence of sub-test results but divergence. Then one could obtain a useful cognitive profile, not a virtually meaningless, so-called IQ.

Re: “those correlations aren’t all that small”

Looking at this correlation matrix the average correlation coefficient appears to be around 0.3, giving a mean r squared of less than 10%. is that data set so atypical?

I guess the question is “atypical compared to what?”

The Woodcock-Johnson battery you linked to seems to test a variety of things. Check out the intercorrelations for test #11 sound blending for example. There are some subgroups of the tests which have higher correlations. For example, look at the correlations for #13 Oral Vocabulary with tests 20-28.

This paper gives somewhat higher intercorrlations for the WAIS subtests (Table 1): https://www.sciencedirect.com/science/article/pii/S0887617797000152
Overall they advocate a 2 or 3 factor solution, but AFAICT did not give quantitative info to judge how necessary that was. They mention tests used, but don’t give the actual data. How hard would it have been to include a scree plot to let readers judge for themselves?

Overall I think the correlations of the individual tests with g is a better measure of g’s significance than the test intercorrelations.

However, correlations among intellectual capabilities are not that good, e.g., between SAT math and SAT verbal, for example (as exemplified by a niece that the Wizard mentioned, who as I recall was a verbal wiz and a math clutz — relatively, anyway).

Table 4 (page 18, 23/26 of the PDF) of https://www.ets.org/Media/Research/pdf/RR-99-02-Dorans.pdf
gives the SAT math/verbal correlation as 0.71 which seems quite high to me (just over 50% of variance explained). Especially when considering the correlation between the SAT Verbal and ACT English or Reading is 0.83 (i.e. not that much more).

I think much of the clutz/wiz dichotomy people observe comes from two effects:
1. Tendency to specialize. This both results in a practice effect and means in most fields asymmetric skill profiles are common and people spend the most time with others in their field.
2. Restriction of range and selection effects. For example, people tend to be admitted to college by some measure of overall ability (e.g. combined SAT). Think about two colleges with admission thresholds of 1400 and 1500 SAT. If everyone goes to the best college they can then the first college has a group of 1400-1500 scorers. The people who did best at both will end up at the second college leaving the first with a larger proportion of unbalanced scorers. This is most easily visualized with a scatterplot of the two correlated 0.71 variables with the thresholds overlaid.

Thanks for your comment.

Re: “those correlations aren’t all that small”

Looking at this correlation matrix the average correlation coefficient appears to be around 0.3, giving a mean r squared of less than 10%. is that data set so atypical?

Re: “I find it quite plausible that various aspects of intelligence all correlate with drivers of good metabolic functioning”

If the cellular substrate of neurological activity, whether functioning well or badly, is perfectly uniform within the population then it seems to me that g might well be zero. If not, then g would have to be attributed to correlated within population variation in the functional effectiveness (as measured by IQ-type tests) of different brain modules. To some extent this can be expected, for example, as a result of assortative mating. However, correlations among intellectual capabilities are not that good, e.g., between SAT math and SAT verbal, for example (as exemplified by a niece that the Wizard mentioned, who as I recall was a verbal wiz and a math clutz — relatively, anyway).

I wonder, incidentally, whether that is not common in girls (I know there are first rate female mathematicians) but in the mating stakes verbal reasoning likely counts for more than math (cf. Jane Austin – Pride and Prejudice).

A further complication is the reassignment of neurological resources among brain functions, for example, the reassignment of the visual cortex to non visual functions in the blind. Such reassignment may perhaps also occur as consequence of intensive use of particular faculties. In that case ability is raised by use: at the expense of loss of capability of less intensively used faculties.

That use, practice, training, etc., increases intellectual capacity seems almost certain in view of the Flynn Effect, and means that correlation among facets of intelligence will never be perfect.

There are several reasons why r-square is preferable to r when talking about goodness of fit instead of correlation r. The most important one is that variances are additive and thus the total variance of data is the sum of explained and unexplained variances. So if r^2=0.5 then 50% of variance is explained and 50% of variance remains unexplained. There is no law prohibiting statements like this: correlation=sqrt(0.5)=0.707 implies that 70.7% of standard deviation is explained. This statement may boost the mood of the modeler nut it misleads because it obscures the fact the unexplained standard deviation is not 29.3% but also 70.7%. Explained and unexplained standard deviations do not add up to 100%.

While it is true that slope=correlation when Y and X data are normalized to the same variance, I do not think that outside some narrow circles anybody would recognize that slope is a measure of a goodness of fit. It is rather correlation and its square that every student of statistics will recognize as measures of goodness of fit somehow linked to residuals from the fit.

There is however a case where a slope is directly related to variance explained. The slope in breeders equation is equal to lower case heritability h2 which in turn is related to some variance explained (by genes). BTW, while I have seen a proof that ∆Z<S and thus there is a coefficient k<1 that ∆Z=k*S but I haven't seen a rigorous proof that this coefficient k actually equals to lower case heritability h2.

Yhanks. So when we compare that low predictability with the high estimates of heritability (whether 40 or 80 per cent or in between) it would appear to follow that we should expect to find many times the so far known IQ related SNPs, alleles or whatever, plus perhaps many relevant interesting epigenetic phenomena…….

I think CanSpeccy’s original point was sound. I see your point more as a philosophical observation about how to define “component.” As an example, I think a persistent intense headache can impair intelligence. Being headache free is IMHO important to intellectual functioning. But for that case I agree calling it a “component” would be inappropriate. I think the cardiovascular system is intermediate. It would be interesting to see how measurements of CV functionality (perhaps stroke volume normalized by body size?) correlate with IQ.

Put another way, being important for the functioning of intelligence and being part of a concept (say measurement or definition) of intelligence are arguably both “components” of intelligence. But they are clearly different and being clear about which is intended is helpful. This sort of multiple definitions become less useful when people use them to conclude “intelligence” is not a meaningful concept.

What kind of prediction does it allow us to make?

Please reconsider your description of the cardiovascular function as a “component” of general intelligence. One might as well say that lung and digestive functions are also components. Reasoning is a (very big) component of intelligence and a good case can be made for saying that intuition is. But physical preconditions for the functioning of intelligence are a different kind of thing – not components. I think it matters to your argument.

Thanks for laying that out in detail. I think this one is problematic. Especially given the research available which I think indicates those correlations aren’t all that small.

(c) Given the genetic and developmental independence of the modules comprising the brain, one would not expect, and does not observe, close within-population correlations among measured cognitive capacities.

For an analogy, consider the quality of buildings within a city. There will be variation among the buildings, but also substantial correlation based on the typical quality of concrete, steel, laborers, etc. available in that city. I find it quite plausible that various aspects of intelligence all correlate with drivers of good metabolic functioning (e.g. myelin production).

I am not sure of the exact provenance of that spreadsheet (can anyone elaborate?). In the Infoproc comment nooffesnsebut states: “here are the graphs of average polygenic scores for the recently leaked SSGAC education attainment SNPs” and “the correlation from Lee et al, using 2985 SNPs, is 0.90″ which leads me to believe they are looking at a still unpublished study from James Lee et al. using the SSGAC data. Probably an iteration of the work in this talk: http://programme.exordo.com/bga17/delegates/presentation/214/

Which I think is an iteration of https://www.nature.com/articles/nature17671
The figure below indicates that study used SD units for the effect sizes and gives an idea of the effect size magnitudes seen in that earlier study.

The big problem I see with the SNPs in the recent spreadsheet is that they don’t distinguish independent SNPs (as defined by LD). The first five SNPs (smallest p values) provide a good example:

MarkerName CHR POS A1 A2 EAF Beta SE Pval
rs9859556 3 49455986 T G 0.6905 0.029 0.001 3.98E-91
rs7623659 3 49414791 T C 0.3095 0.029 0.001 5.49E-91
rs11917431 3 49644012 T C 0.6973 0.029 0.001 6.86E-91
rs1873625 3 49666964 A C 0.6973 0.029 0.001 7.52E-91
rs11921590 3 49644193 T C 0.6973 -0.029 0.001 1.12E-90

Those all look like essentially the same thing (especially the third and fifth which are only 180 nucleotides apart, note that the graphic above does account for independent SNPs per the linked letter). I don’t know how they computed betas and/or a prediction function,

One thing I do not understand is how the third and fifth SNPs (rs11917431 and rs11921590) have the same frequency (EAF) but opposite signed betas. That leads me to believe there is not a simple mapping between effect allele and A1 or A2, but it would be good to hear from someone who knows for sure.

At this point I don’t think I understand the data well enough to work with it effectively.

It might be interesting to compare the SNP lists in http://www.nature.com/mp/journal/v21/n6/extref/mp201645x4.xls

P.S. I think there are a number of problems with talking about 1500 IQ humans.
- “IQs” that high aren’t really meaningful (well defined) since they are defined by frequency (deviation IQ) and the population is too small. Rasch scores aren’t much better. It’s just not possible to norm an indirect measure like IQ so far into a range where no one exists.
- It is unlikely the linear extrapolation works out that far. For example, there are physical limits on nerve conduction velocity.
- I think envisioning a world where John von Neumann is average and there are people 4SD smarter than him (which I do think is achievable) is sufficiently challenging.
- Sensationalist speculation like that can make IQ research look less serious. Perhaps it was good for getting attention, but not a good angle to focus on longer term to be taken seriously IMHO.
It does make a great headline for those who like sensationalism though.

A lot of wild conjectures about “education”. We still don’t know, we don’t have any proven imprescindibility of “education” to increase human intelligence. In the end of day we already live at least in developed countries a educational system which is considerably equitable and the results don’t change only with demographic changes via migrations. Even in countries as Brazil we already have a relatively equitable educational system. But almost things school “teach’ is how to memorize a lot of abstract information, or, to try to memorize.

Your last paragraph many people here already know. Are you suggesting almost people don’t know about this?

But… In arts at least specially in literature verbal ability/ies tend to be hugely required and in visual arts non-verbal abilities also are very required at least in non-pseudo-talent, and visual spatial abilities IQ tend to be good to measure and predict. Again you are too far from reasonableness when you starts to deny even the basic facts about IQ. While I’m with you against IQ fetishism I’m against you when you act just like Socjusticeworriors or a useful “idiot’.

In entertainment at least to be a host in “very important’ tv show you just need have verbal agility + psychological factors as charisma. But I’m not saying this cognitive factors (IQ is good to measure, I said GOOD) are extremely important because we have a lot of uncharismatic, shy or whatever non-socially skilled people who score higher in verbal abilities, I mean, mechanicist verbal abilities (mechanicist or cognitive). It’s always mostly a combination of, specially, intrinsic factors which predict given potential and not just OR fundamentally IQ, what most IQist believe. Of course I’m talking about potential prediction but not subjective or not soo subjective factors as luck to explain most of very famous host tv show. Even to be humorist many components of verbal mecanicist cognition (something IQ measure) are required. IQ analyze the ingredients and not the recipe.

We need start from something and IQ has been good for this function, to start.

Mathematical, verbal and spatial seems very multifaceted even because as you know verbal and maths already tend to have their mutual modular tilts.

Do you believe IQ 70 today is always related with mental retardation?? I mean, pre historical humans were “retarded”??

To be even more explicit, what I am saying is this:

(a) The brain is both structurally modular — as the neuroanatomy shows, and functionally modular — as the neurophysiology shows.

(b) Every brain module has its own genetic determinants and its own developmental history, the latter subject to environmental effects which are generally of great importance, for example, input from the optic nerve during development of the mammalian visual cortex is essential to the acquisition of vision.

(c) Given the genetic and developmental independence of the modules comprising the brain, one would not expect, and does not observe, close within-population correlations among measured cognitive capacities.

(d) Despite the genetic and developmental independence of the functional components of the brain, all depend on the same cellular constituents — mitochondria, membrane lipids, pumps, channels, protein kinases, the machinery of energy and intermediary metabolism, and much else beside.

(e) Variation in this cellular substrate of neurological activity means that there will be common factors affecting the functionality of all components of the brain. This is what gives rise, in part, to the modest, ca 10%, population-wide regression coefficient among cognitive capacities, i.e., the g factor. In addition, there is an environmental component to g, including nutrition, education, and much else beside.

(f) Intelligence, being multi-faceted, cannot be measured with a uni-dimensional scale. Moreover, variation in relative capacity in different cognitive dimensions will be heavily influenced by cultural and other environmental factors, hence, the Flynn Effect.

(g) Notwithstanding (f), uni-dimensional measures of cognitive capacity in limited domains may have power to predict such things as academic achievement, or career prospects. This is not surprising since (1) career success/income tends to be reflective of verbal or mathematical reasoning capacity, and (2) education focuses on the development of verbal and mathematical reasoning capacity, and (3) so-called IQ tests (therefore) focus on verbal and mathematical reasoning capacity. Such tests are not, however, necessarily useful in predicting success in many other domains, for example, athletics, the arts, and entertainment, or indeed in predicting mere survival in contexts other than a developed economy.

Indeed, .265 is approx the square root of 0.07. I’ve always had a bit of trouble getting my mind round variance as the proportion of something explained by whatever.

If one is being told here that 7 per cent of the difference between the measured IQs of two test subjects is, on average, explained by what is known about (quantities associated with) certain specified genes that seems to be very little, especially given the probability that genes of major effect would be discovered first. Moreover the great lump of genes and other causal elements which explains the other 93 per cent might hypothetically have effects which are both positive and negative, all over the place indeed.

res, I am trying to be careful with these extra carriage returns. I am also not sure of the identity of the bolder either. The tone of the initial rebuttal surprised me. Can there be any doubt that a substantial shift in human IQ is now on the horizon? Probably should be more ranting and raving than there is. I am doing my part to fill in this gap. Is it not totally clear to everyone that we are rapidly approaching a Singularity? It is almost pointless to even debate this anymore. We are truly staring ahead at this social chasm. Sure, it might not be 1500 IQ people, though it does not
need to be anywhere near that for massive social disruption. So many of the unz threads go on endlessly about a few IQ point differences between groups of people. The differences that we now have on the table are no longer a few points.

You noted, 40 IQ points –> ~1 PGS score. So 1 “SD” = 40 IQ points. As a guess, it is simply a scaling question. The true scale on the y-axis should have a range of 3 when all the SNPs are found, but they have only found a third of this so far.

Below is the list of the 5 most positive and 5 most negative SNPs. You were right. The ones that were negative were almost all rare, while the positives were mostly common. It was striking, though, that so many of the large absolute betas were not genome significant. Look at the big positive ones below: 0.995, 0.997, 0.997. These SNPs are going to fixation and yet have not quite made it. It will likely be very helpful to call in other populations to help resolve some of these. As I said above, this size GWAS has the power to detect genome significant SNPs around 2-5% tails with absolute betas of 0.02 or greater.

MarkerName CHR POS A1 A2 EAF Beta SE Pval
rs137861526 13 63818382 T C 0.005102 0.1 0.024 2.38E-05
rs6995160 8 58608236 A G 0.003401 0.088 0.02 1.67E-05
rs185897402 2 153507714 A G 0.994898 0.079 0.018 8.29E-06
rs146409194 7 66040951 A T 0.996599 0.078 0.018 1.17E-05
rs117899362 12 49937297 T C 0.996599 0.073 0.016 9.75E-06
rs190490932 6 27236857 T C 0.996599 -0.074 0.017 9.23E-06
rs77414068 1 96274765 A G 0.003401 -0.075 0.017 1.57E-05
rs188424966 7 106564157 T C 0.001701 -0.076 0.018 3.83E-05
rs4684749 3 10996283 T C 0.005102 -0.08 0.02 4.15E-05
rs190410522 2 58389547 T C 0.001701 -0.099 0.022 6.66E-06

I was able to write a program that extracts the genotypes from my genochip results for over 500 of the 3000 SNPs. I have written another program that uses proxies to see how many more of the SNPs I can account for. The one thing I am not sure about now is which of the SNPs is the effect allele. Look up the first two SNPs in the table above on dbsnp. The effect allele shifts back and forth! Would it really be required to look up the frequencies in dbsnp? That would make things quite a bit more complicated for my program. Couldn’t they have used a simple rule, such as A1 is the allele associated with the beta?

Even if EH’s comment does not contain a proper argument, it at least has the virtue of being correct. Your comment on the other hand has neither.

The bold text is typically the moderator. In this case presumably Dr. Thompson (though I am not sure about that).

Notice how in comment 253 there are carriage returns after each sentence (they were not removed in that comment). In long comments having those returns makes it harder to read. Best not to use the carriage returns like that. I often go to considerable trouble to remove extra carriage returns from text I cut and paste from papers for the same reason.

Your wild assertions run contrary to all the research, of which you are entirely and deliberately ignorant. Your foolish posts are a waste of time even to read, so arguing with their maliciously ignorant author would be even more so.

This is not a rebuttal. Even in high school you should have learned how to critique a document. You get a big fat F.

Once again, another IQist proves to be a complete idiot.

Or to be more explicit, how could cardiovascular function not be a component of general intelligence when the brain is by far the most energy intensive organ.

In the case of cardiovascular disease one can see the effect on mental function. I recall, in particular, a man of around 60 who was both slow of speech and dull of mind: apparently a case of early onset dementia. But after aortic valve replacement surgery and a long recovery period, he regained normal speech and the full mental vigor of a man of middle-age.

Other general physiological factors affecting intelligence are not difficult to envisage.

[You are NOT supposed to press Return!!]

What am I doing wrong?
I write a sentence and then I press return.
How could that be my fault and not the forum’s software?

Maybe.
Maybe you no have more arguments..

So it’s matter too*

I think you’re too much against IQ and become a opposite of IQist in the sense you deny most of IQ stuff but not in essence, in the sense you are too radical and less balanced/less wiser [as a typical IQ-fetishist] to approach this matter without be out of reasonableness, i mean, always required one.

No. They are components of human intelligence. There are many other both cultural and cognitive elements of human intelligence.

I never deny this. And not, they are not JUST ANOTHER components of human intelligence, it was what you mean.

Right.

I don’t think so.

May be you’re suffering a brain oxygen deficit.

But language and mathematics are the cultural and cognitive basis of human intelligence.

No. They are components of human intelligence. There are many other both cultural and cognitive elements of human intelligence. You have to watch out or you’ll find yourself squarely in the IQist camp.

determines brain oxygen and glucose supply), and other such features of organismal physiology, without a specific relationship to cognition.

I don’t think so.

In general, I think Winston Churchill was correct in saying that genius is one percent inspiration, 99 percent perspiration.

Edison, not Churchill: http://www.quotationspage.com/quote/2054.html

Without 1% of inspiration no have 99% of perspiration or transpiration.. Think about it.

But language and mathematics are the cultural and cognitive basis of human intelligence.

[Once again: stop putting carriage returns at the end of every sentence if you want your comments published.]

Sorry, not sure what is causing that.
Whose text is in bold?

oops, the “despite an IQ too low to measure, was supposed to refer to Derek Paravicini, not Winston Churchill, who despite being an academic failure, was not a moron in any way.

I can assure you the correlation between IQ and e.g. arithmetical speed and accuracy tests or who put their hand up first to offer a translation of a Latin text was very high. So I say that there can be no doubt “speed of processing” by whatever name is indeed “an important nervous system variable”. Indeed, on reflection I wonder how you could deny it without being an anti science lefty PV troglodyte which you do not appear to be.

IQ measures, mainly, verbal reasoning and mathematical reasoning ability. So of course it correlates with academic achievement in those subjects. More generally, achieved ability in math and languages is going to correlate with other aspects of academic performance, if only because it indicates motivation to achieve academic excellence, excellence that is in every test that school has to offer. You must know people like that, driven to compete. Others strive to drive holes in one, like Moe Norman, or run miles under four minutes, or get a lot of money, or whatever.

As I’ve acknowledged above, IQ tests correlate with academic performance. But academics are really only a small part of life. My big sister, who had to repeat a year at the grammar school, became a fine concert pianist, my other sister, who has something like an eidetic memory and won a big scholarship to university, never accomplished much of anything, though she undoubtedly has excellent logic circuits and what you might describe as a quick mind.

In general, I think Winston Churchill was correct in saying that genius is one percent inspiration, 99 percent perspiration. Churchill, who flunked everything but English in school, and remained in the lowest form at Harrow for three years, worked endlessly at the construction of the plain English sentence and brought its use to the point of genius, despite an IQ that is probably too low to measure.

And if by processing speed you refer to some aspect of neural activity, where’s the evidence? Action potentials in myelinated axons travel at up to 100 m sec-1. I doubt if they travel at a very different speed in my brain or yours than in Derek Paravicini’s brain. In fact, if speed of nervous conduction had anything to do with intelligence, I’d guess it might be faster in Derek’s brain than the brain of most, since he is surely a keyboard genius.

One way to sober up is to generate 20 or so (how many countries you have?) random numbers and then searching database for say 100 SNPs that produce polygenic scores that have the highest correlation with these numbers. Davide Piffer actually did it, sadly to no avail, and found SNPs that explained his data set better than SNPs he selected based on some inside knowledge. Among 10 millions of SNPs one can find subsets that can explain any not excessively long random numbers sequence.

I’m nor sure that “speed of processimg” is the precisely correct technical term but I recall reading of high positive correlations found years ago between reaction times and IQ scores. Just from memory I can recall – as surely you can – tests at school which showed just how much faster some children’s brains dealt with material that all were equally practised at than others’. And, as one who had access once to the IQ records of a school I can assure you the correlation between IQ and e.g. arithmetical speed and accuracy tests or who put their hand up first to offer a translation of a Latin text was very high. So I say that there can be no doubt “speed of processing” by whatever name is indeed “an important nervous system variable”. Indeed, on reflection I wonder how you could deny it without being an anti science lefty PV troglodyte which you do not appear to be.

The SNPs with the largest effect sizes were almost all not genome wide significant.

Probably worth checking if those SNPs are low frequency deleterious variants. There probably are some real effects there, but the frequencies are too low to give a good signal. Who knows how accurate the coefficient estimates might be though. Probably getting a lot of noise in those. High frequency large effect deleterious variants seem much less likely due to selection (though potentially most valuable for improving population IQ if any do exist).

It would be interesting to do some analysis of coefficients and frequencies. That’s a pretty common thing to plot, but I don’t know if anybody did it for this dataset.

I wonder if calculating which variants have the largest population effect (effect size scaled by frequency) would tell us anything interesting. I’m curious how well that would correspond to p values. I’m somewhat busy today, but should really spend some time playing with the data in that spreadsheet.

The starting point of this unnecessary excursion was the gap between twin based heritability and GWAS and Hsu’s variance explained. The gap for height is still 80% vs. 40%. Can rare variants (<1%) that presumably were not in Hsu's database close the gap or are critics like Richardson correct that twin based heritability is flawed? I can see that by adding more SNPs and adding nonlinearity to polygenic score the predictor function will get better but at the same time I can see that twin based heritability is overestimated. I think this overestimation is more severe in the case of whatever stands for intelligence than height. So far we have Hsu's 9% for educational attainment and 7% from GWAS studies like in the paper discussed here. And the possibility that stratification may have something to do with these results may not be easily excluded. Actually I do not know how to go about dealing with stratification. Perhaps by using a subset of population where parent's income does not correlate with children educational attainment. If a predictor function on such a subset works worse than on the whole set then it may be an indication of stratification in the data set.

[For some reason, you seem to add a carriage return after every sentence, which requires fixing to avoid having your comments look terrible. Since you mostly rant-and-rave about nonsense like IQ=1500, please stop doing that if you want them published in the future.]

I think we need a thread roll call on this one. All those in favor of a new blog post for this new result say Aye.

Aye. I say the Ayes have it.

This is a tremendously important result. We have been waiting for this one for eight months and here it is! I am not sure how much longer people will be able to bluff anymore. There WILL be substantial sociopolitical disruption over the medium term due to this technology. We do not have all the background info from the article, though I wonder if we have reached the point now where embryon selection for enhancement of IQ is truly feasible.

Sorry, res. Too excited to write clearly.

Yes, this is true about EA only being an OKish proxy, though it is not that bad. We at least are now the right ball park.

And you are also right that adding in all the SNPs (almost 10,000) really is not overly helpful. Probably best to stick with the first 3,000. Yet, I was disappointed that so many of the big ones were then lost.

I have the SNPs in a spread sheet and I am just messing around with them. With “98% with 0.02-0.03 under the p-value line”, I wanted to see where the line was to detect genome-wide significant SNPs ordered by frequency. The line seems to be around the 2% population frequency tails (below 2% and above 98%) for the 0.02 to 0.03 SNPs.

The SNPs with the largest effect sizes were almost all not genome wide significant. As I noted only 5 of the top 100 largest positive SNPs were significant. This would be such a great opportunity to call in other populations (Asia?). Not sure what the ethnic composition of the current sample was, though I guess European.

Yes, I noticed that too about the outliers from the regression. This could show how large environmental effects truly can be. If they could prove that such large effects are possible, then it would take quite a bit of the edge off the whole idea of genetic determinism. I would think that developed nations should make it a priority to be positive outliers on the regression.

Sri Lanka is a large negative outlier (about 1 SD). If I were them, I would get on this research and find how they can move to their potential. 1 SD is large. It could have a substantial impact on their per capita GDP.

But I think you miss the point that it is not being a good soccer player or brilliant pianist that causes people to attribute high intelligence but rather, in complete contrast, the ability to quickly make sense of the unfamiliar**. Speed and a good working memory are of the essence.

LOL. I just go by the dictionary, Websters and Oxford, both confirm my position. IQists can measure what they like, but they are misleading the world by calling some specific faculty of mind “intelligence,” when it bears little relationship with the common usage of the term.

it is speed of processing and the finding of the obscure answer to the surprise question that will count at least as much.

What do you really know about “speed of processing.”

Listen to this. What key’s it in? Any musician, even if they’d never heard it before, would tell you in a flash*. Play him a 6-note chord in C flat major and gamma-minus moron Derek Parvicini will not only give you the key, he’ll hit all six notes on the piano without hesitation.

Lawyers train to give the quick, the surprising, the learned answer to a difficult question. And indeed you acknowledge that that skill is the result of long training as well as thorough preparation in any particular exercise of the skill. Likewise, musicians train to do the things they do (and fast).

If you think lawyers are so smart, try blindfolding one and asking him to instantaneously identify and play a 6-note chord. Then you’ll understand that what you call “speed of processing” in any domain is mainly an acquired characteristic, not an important nervous system variable.
———
*D Major, obviously.

From your Hungarian grandmother?

LOL. I happen to be about the only Canadian without a Hungarian grandmother. No, I got it with a ten-second Google search. But it’s quite good, I thought.

I just tell you to try to answer point by point of my previous comment

Well, I don’t subscribe to this:

I believe human intelligence is unitary in its basis/bottom but become very specialized in the top.

The unity of mind arises through integration of the activity of neurologically distinct brain modules, each with its own genetic determinants and developmental history. That is evident from neuroanatomy, and also from psychometric data.

That’s why SAT math and SAT verbal scores are not well correlated, or why the correlations among all testable mental attributes is low, i.e., a mean r squared of less than 0.1.

Such correlations as exist, and which give rise to the g factor are attributable almost certainly to basic physiological factors such as cardiovascular health (which determines brain oxygen and glucose supply), and other such features of organismal physiology, without a specific relationship to cognition.

I’m not sure how to interpret all of this, but some thoughts.

Using EA as a proxy for IQ makes your extrapolation even more tenuous than usual.

If you want to do an extrapolation, I think it would be good to double check that calculating a score based on all of the EAFs gives a reasonable size (say 0.4 – 0.6).

About 40 points of IQ correspond to 0.9 on the genetic score in the graph. Beta in SD units seems like the logical assumption (regression on Z scores), but hard to reconcile with the 40/0.9 equivalence (~2.7 IQ SDs for less than 1 score point).

I assume EAF is Effect Allele Frequency. I find it hard to interpret your sentence: “Large effect betas (0.02-0.03) over 98% EAF are under the p-value line”

US Blacks are an interesting outlier–far overachieving their genetic score in the graph. I wonder what that means. It is odd that US Blacks on average admixed 80/20 with whites have a genetic score comparable to the African tribes.

Jamrs Thompson – would you please explain what exactly “we used out metanalytic GWAS data to predict almost 7% of variation in intelligence….” tells us? What does that equate to in terms of correlations? What does it say about the accuracy of prediction? Does it mean that the null hypothesis in every case is IQ 100 but the metadata allows building on that to say that …. well what?

I didn’t understand enough of the jargon to assess what you wrote but on the graph at the end I noted a surprisingly large gap between the Carribeans and the Puerto Ricans. Any explanation? Anyone?

I see you have had a lengthy dispute with James Thompson but it is right here that I think you have gone wrong. Of course you might have a tribal dialect that contradicts my view of common usage. But I think you miss the point that it is not being a good soccer player or brilliant pianist that causes people to attribute high intelligence but rather, in complete contrast, the ability to quickly make sense of the unfamiliar**. Speed and a good working memory are of the essence. Even if counsel’s brilliance in a superior court, getting an unexpected answer from an exprt witness or a difficult question from an Appeal Court judge will mostly be a result of endless preparation – on the particular brief but also generically from the follow up to or even preparation for his first law school lecture – it is speed of processing and the finding of the obscure answer to the surprise question that will count at least as much. And the more intellugent counsel is more likely to be able to reply confidently to the judge “With respect Your Honor I would answer that with the suggestion that the question embodies a category error” and be able to back it up. Of course I am not denying the contemporary aphorism about ten years to achieve mastery though I do suggest that a healthy person with an IQ of 150 might achieve it in 5 years of standard working days (i.e. assume comparable sleep requirements for all).
** I was about to add politicians as a test case except that I think the 140+ IQ only becomes really important if he/she is a short sleeper with immense stamina.

Very startling!
Here are some of the numbers.

The sum of the positive betas is 18.544.
The sum of the negative betas is -17.062.

If these betas are measured in IQ SDs (could anyone check whether this is true?), then we have:
37.088 SD of IQ for the positives and 34.124 SDs (absolute) for the negatives.
71.212 SD total. Double it for two good (or bad) gives 142.424 SD.

Anyone interested in 2136 IQ?

There are also another 7,000 SNPs reported that are below the genome wide significance line.
Sum of the non-significant SNPs:
37.379 positives and
-36.573 negatives
Double absolute values is 147.9 SD.

219.1 SD 3286 IQ (upper bound with all SNPs included).

Most of the large effect betas (positive and negative) reported were not p-value significant because
they were either very high or very low frequency.

Could try different populations to sort this out.
For example, the largest positive effect size was for rs13786152613

rs137861526 13 63818382 T C 0.005102 0.1 0.024 2.38E-05

C allele frequency for Europeans is 0.006 and South Asian is 0.021

Large effect betas (0.02-0.03) over 98% EAF are under the p-value line
Large effect betas (0.02-0.03) under 2% EAF are under the p-value line

Of the top 100 positive effects SNPs only 5 are genome wide significant at 5×10-8

The figure from infoproc suggests to me that Asian IQ has been underestimated (perhaps due to environmental factors). If you move shift the Beijing Chinese results to the right on the regression line than the polygenic IQ score of 0.7 converts to an estimated IQ of 110.4
(y=0.07x -1.5848; y~0.7)

Your wild assertions run contrary to all the research, of which you are entirely and deliberately ignorant. Your foolish posts are a waste of time even to read, so arguing with their maliciously ignorant author would be even more so.

From your Hungarian grandmother?

Don’t think we’ll ever see a useful test for wisdom.

Wisdom is generally assessed retrospectively.

That is because it concerns questions that are too complex to judge on the basis of mere calculation.

Another huge problem of IQ is that it’s works relatively well when it’s measure and compared great number of individuals but say little about individuals.

I disagree in both three statements you did.

I thought it’s perfectly possible to create a good test to analyse, rank, compare and measure wisdom or WQ. {Pay attention to the difference between ranking and comparison, IQ rank more than compare/part of analysis].

Wisdom is not just by chrystallized intelligence or even fundamentally about experience, it’s also and considerably speaking about fluid skills [don't confuse with ''non-verbal'' skills].

There is what Wisdom ”is’ and how people can reach it. You can do very good judgments without previous knowledge, just with the information you have in your hands. It’s just like a magic cube, always find a way to reach the reasonable or balanced judgment, the best of all, if everything exist fundamentally exist [and specially for long or indetermined term] because it’s in balance.

Wisdom in my view can be used to both, complex and not so complex issues [or even very simple], the big problem of most people starts from their incapacity to accept that they commit mistakes, often frequently, in thinking and in action, specially moral ones. So many people can reach balance in their judgments just because their egos or instincts which blind them to find a path of universal ideality.

What i said about frequency of intelligence versus intensity/size of intelligence. One of the big difference between both is in self knowledge. When i said ”i don’t know, or i’m not totally correct about this… i can be wrong”, i’m being more smarter or better, wiser, than when i try to ”rationalize” or self-deceive me that ”i’m perfectly capable to understand or to explain this” and find a way to show my size-intelligence/frivolously intelectual creativity. So this frequency or secrecy would happen all the time, already in the micro-level of our lives, from our homes to the big events or ”in the streets”, in interactions with another peoples and circumstances. Learning from the micro to the macro level or at least try to reach all levels, but always starting from the micro.

res, it’s Christmas in March!
Best Day ever!

A comment on infoproc mentioned that the huge Educational GWAS results have finally been
leaked. Almost 3000 SNPs. Wow. The IQome has been unlocked!
Anyone know how much variance this explains (guess we could always sum up the betas).

Humanity desperately needs to talk about what is happening now.
We need leadership to help the discussion that needs to start.
The journey to eugenica has started.

Argument over.

https://drive.google.com/file/d/1gF9a4oSvevU4gdbuRmDEbMZwIRiSHVK-/view

When I have a translator in computer I will answer your comment above. I just tell you to try to answer point by point of my previous comment, don’t believe it’s a offense, just to this debate be more fluid and not with this abruptivity.

Hbds and psychometricians always confuse outliers with outsiders. Geniuses are often or characteristically both. It’s not all outsiders who are brilliant. Indeed to be outsider you must have some disorder to be out of social evolutive chain and have a different existential and perceptual perspective. Outliers are those who are too beyond the average but still in-the social evolutive chain on the “right” or “left” side of bell life curve.

A already provided here some times a proposed differentiation, qualitative differentiation among gifted people: high achievers, brilliant learners and creative thinkers. My belief is that IQ can access well the cognitive properties of the first group and less than the two last.

SNP could be so redefined as to encompass all other changes.

Or you could just use existing terminology which is well understood by people in the field. I’ll leave it as an exercise for the reader which approach usually works better.

Feynman talked about lesson he got form his father lessons of not being limited by names and classification.

This is actually a good thing to bring up. Striking a balance between doing things your own way and basing understanding on prior work can be difficult. IMO going your own way is much more difficult and best reserved for cases where one has exceptional knowledge, ability, and willingness to put in a great deal of time and energy. It also works better if one happens have Feynman (or Feynman pere most likely) level ability.

There is a school of thought that always doing things your own way fosters an innovative mind, which I think is where you are coming from. I have had very smart coworkers who always tried to take a cut at a problem before looking at other people’s approaches. There is something to be said for that (I try to follow that example to some degree, depending on my level of background in an area), but knowing how big a bite to take at a time can be challenging. I am pretty sure they did not try to reenvision entire academic fields (e.g. genetics) at once. I tend to prefer being innovative in how I make connections between bodies of existing knowledge rather than building everything from scratch. I am not Feynman and to pretend that I can do a better job of building all of the intellectual infrastructure (the work of thousands of smart people over decades) of one of my interests (here genetics) to the point where I can even begin to do new work is beyond hubris IMHO.

Put simply, I prefer breadth over depth because I think multidisciplinary work is underemphasized and there are more opportunities there. It also allows me to seek out areas of interest and current potential. This requires reliance on domain experts accompanied by the testing of their conclusions. If you have read my comments here, do you really think I lack the willingness to challenge experts in their own domain? The problem is, to do that productively I must be able to communicate in their language, not make up my own. It also helps to make an effort to be polite and humble (e.g. always realize I might be the one making a mistake). I struggle with that sometimes (as is probably obvious to people who read my comments), but I think you are to the point of being a good cautionary example of what not to do.

The funny thing is, growing up I had much more of a tendency to come up with my own ways of doing things than most people around me. Works better when one is much smarter than most of the surrounding people. Being like that becomes infinitely harder when one enters much more selective and competitive worlds.

Pls to do try to obfuscate issues with your BS of superior knowledge of things that are really irrelevant to the gist of the idea.

That is an odd objection. My point was about non-SNP heritability and I described what non-SNP meant in some detail. Highly relevant. And I’ll leave it to others to decide on the balance of clarity vs. obfuscation in our respective comments.

You know exactly what I have meant and your objections really came from ill will.

It must be nice to be absolutely certain of other people’s motivations all the time. When I see responses like yours my gut reaction is “Projection!” Hard to be sure if that is true or not, but it seems the most likely explanation to me on reflection. One tip, if you are going to be an innovative, Feynman-esque wonder do NOT assume other people always understand you. If anything, the opposite is more likely to be true.

As far whether the idea that rare variants in large number can account for trait distribution in any significant way one must look at distribution of polygenic score consisting of say 10,000 rare (<0.01%) SNPs. This polygenic score may account for wings of the distribution but not for its main body within ± 2SD.

Hard to say for sure (I’m not even sure if your last sentence is meaningful). One way to think about this is the idea of genetic load. It is important to remember that I am talking about explaining the missing heritability here. The numbers we see from GWAS etc. make it clear that much of the heritability is simply additive SNPs.

As far as getting educated I will pass on your suggestion. Too much learning dulls one’s intelligence and ability to think independently. I do not want to deplete already limited amount of intelligence I am left with in my old age.

Unsurprising, but still fascinating. Perhaps you could talk about some of the intellectual successes you have achieved with this approach?

SNP could be so redefined as to encompass all other changes. Like for example MNP or CLUMPED which are really sequences of adjacent SNPs but since they are adjacent we do not call them SNP. INDEL also can be seen as an SNPs. Here it is how it is done now but it does not have to be so:

https://genome.sph.umich.edu/wiki/Variant_classification
1. Trim each allele with respect to the reference sequence individually
2. Inspect length, defined as length of alternate allele minus length of reference allele.
1. if length = 0
1. if length(ref) = 1 and nucleotides differ, classify as SNP (count ts and tv too)
2. if length(ref) > 1
1. if all nucleotides differ, classify as MNP (count ts and tv too)
2. if not all nucleotides differ, classify as CLUMPED (count ts and tv too)
2. if length ≠ 0, classify as INDEL
1. if shorter allele is of length 1
1. if shorter allele does not match either of the end nucleotides of the longer allele, add SNP
classification
2. if shorter allele length > 1
1. compare the shorter allele sequence with the subsequence in the 5′ end of the longer allele
(count ts and tv too)
1. if all nucleotides differ, add MNP classification
2. if not all nucleotides differ, add CLUMPED classification
3. Variant classification is the union of the classifications of each allele present in the variant.
4. If all alleles are the same length, add MNP classification.

Feynman talked about lesson he got form his father lessons of not being limited by names and classification.

The bottom line is that when constructing any predictor function you include in it everything what is there, i.e., all differences with respect to a reference DNA to which you can assign any value of numerical trait as an offset. This is what you start with and then you look for the function with minimum number of parameters just like Hsu.

Pls to do try to obfuscate issues with your BS of superior knowledge of things that are really irrelevant to the gist of the idea. You know exactly what I have meant and your objections really came from ill will.

As far whether the idea that rare variants in large number can account for trait distribution in any significant way one must look at distribution of polygenic score consisting of say 10,000 rare (<0.01%) SNPs. This polygenic score may account for wings of the distribution but not for its main body within ± 2SD.

As far as getting educated I will pass on your suggestion. Too much learning dulls one's intelligence and ability to think independently. I do not want to deplete already limited amount of intelligence I am left with in my old age.

Don’t think we’ll ever see a useful test for wisdom.

Wisdom is generally assessed retrospectively.

That is because it concerns questions that are too complex to judge on the basis of mere calculation.

For example, when Hannibal decided to take elephants over the Alps, how many of his commanders said “hey, good thinking, not.” Several, I should think.

And if something had spooked the beasts in transit and they’d stampeded over a cliff, the nay sayers would have been proved right.

Likewise, with Elon Musk’s plan for the colonization of Mars. If the first ship with 100 on board blows up on the launch pad, the second ship with 100 on board crashes on landing at Mars and the third with 100 on board blows up during take-off for the return journey, Musk’s reputation for judgment will be at zero.

But if the first party arrives at Mars safely and returns home safely, then Musk could run for President, and win.

res, very exciting!
We had the g factor and now we have the p factor!

When are the grown-ups going to get this and start putting in the serious resources
that the p factor GWAS’ deserve? If psychopathology is psychometrically unitary, then a utopian
world without “differently normal” might be on the horizon. The social investment returns of the genetics of psychopathogy just got much more bullish.

https://www.biorxiv.org/content/early/2018/03/23/287987

There are many aspects of intelligence that the IQist ignore. What their tests provide, primarily, is an assessment of verbal and mathematical reasoning ability.

Seems obvious to perceive.

The problem with the IQists is a lack of PQ (philosophical quotient).

Bingo!
What i said some times here: IQ is great to find the best WORKERS but not exactly the best HUMANS/BEINGS, i mean, in our totality.

They evaluate people for several abilities that can be assessed with a quick paper and pencil test, and call that a measure of intelligence.

Yes, they call cognition = intelligence and they even don’t believe in the existence of emotional intelligence, what a flock, humans as well maybe most of living beings, behave in emotionally-smart ways. JUST because emotional intelligence has been used, racked by leftoids to attack IQ and racial differences, IQists decided to act in same way, so childish and dumb, jee!!

So, they separated intelligence [something simply don't exist in real world] from emotion and call only cognition as intelligence.

Then, based on those results, they grade people on a linear scale from moron to five-star genius, while ignoring that the moron may have the musical genius of Derek Paravicini, or the visual and artistic gifts of Stephen Wiltshire, whereas the five-star genius may be incapable of carrying a tune, recognizing a face, or cutting a two by four square.

Yes, but you need less extreme exception-nal examples to be used because may exception prove the rule in this case and i don’t believe.

I’m perfectly fine to accept the genius is not even positively correlated with some very important features almost people desire, regardless for what, and most of this set of features are basically of wisdom. We have mental disorders, and i believe wisdom is a kind of ”extreme’ mental ORDER. Wisdom is the frequency of intelligence, ”intelligence”/IQ is the size and at priori intensity.

The problem is that they OVERVALUE what they believe it’s all conceptual and practical totality of intelligence as a incomparably good thing, on other words, they don’t care about evil geniuses and specially those who have the same lower moral levels than many them.

IQ is all about size, intensity of cognitive features, a very important part of intelligence, it’s quantitative, but in the QUALITATIVE aspects they not just are far to be knowledeable but seems they don’t want accept them/quality of intelligence.

The IQists justify this presumption by claiming that their tests measure a unitary property of the mind, which justifies their fascistic claim to be able to grade all of humanity on a uni-dimensional scale, thereby establishing a global pecking order for once and all.

You must need separate what is a fact and what is a value. Based on what IQ reach i don’t believe they are factually wrong about that, also seems a obvious thing, but the problem is the intentions, and specially the hidden intentions.

But the evidence, neurological, anatomical and even, psychometric, refutes their claim. Intelligence is not a unitary property of mind and, therefore, IQ is a misnomer for whatever it is that the IQist’s measure. Moreover the use of the term IQ involves an egregious abuse of millions of people subject to the IQist measure, since it labels them in a way that in no way reflects the multiplicity of their natural talents.

I believe in three scenarios

a species where evolution promoted only one type/unitary of intelligence.
a species where evolution promoted a diversity type, only-one and combinatory ones
a species where evolution promoted a hyper-specialization within itself or just higher [psycho] cognitive diversity.

Human scenario look like the second one. And also, i believe we must need differentiate what is basis and what is not. I believe human intelligence is unitary in its basis/bottom but become very specialized in the top.

Yes, i think seems extremely important we have analytical and deep support in psychometric research, because just the psychometric results don’t appear to be enough to give a complete analysis about people, at individual levels and having reverberations at collective levels.

As RR88 points out, the IQist’s notion of intelligence has no “construct validity,” nor therefore, any operational definition. Instead, reliance is placed on a circular definition. IQ tests measure intelligence, therefore, intelligence is what IQ tests measure.

I believe no have validation and be conceptually partial is two different things. Yes, they are very lazy and incorrect in this aspect even it’s appear to be a strawman, it’s something IQism is in its very basis and it’s practiced routinely, intelligence concept was or has been replaced by IQ in many psychometrician works.

If Psychometricians would abandon their arrogant pretensions, and acknowledge that what their tests attempt to measure are specific abilities, these the product of both genes and environment,

Well, but if at least IQ measure some ”specific abilities” so i think it measure something and it’s valuable in the end of day. But i don’t believe verbal and mathematical are SPECIFIC as music ability really is, but a very basicl/primary capacity which represent the human intelligence foundation.

What i already said here

we have: creative, analytical, memory abilities, basically. We perceive/analytical, we internalize/memorize and we create/invent/combinated. IQ measure/and compare very well the late: memory, specifically semantic memory//general knowledge. About analytical ability, IQ measure in: non-real world context or very superficially. About creative ability, it’s what we already know, IQ correlates accidentally with it because there are some overlapping between GENERAL cognitive intelligence and creativity. My opinion is that highly creative people tend to have great cognitive amplitude between their abilities and because IQ is mostly centralized in generalities, it’s missed.

Iqists hate;
don’t understand;
don’t care about creativity

There are many aspects of intelligence that the IQist ignore. What their tests provide, primarily, is an assessment of verbal and mathematical reasoning ability.

Such tests are not without value. If I were the war minister of Ruritania with 100,000 fresh conscripts to deal with, I’d probably subject them to a so-called IQ test and assign the top 5% for officer training and the bottom 5% to peeling potatoes.

The problem with the IQists is a lack of PQ (philosophical quotient). They evaluate people for several abilities that can be assessed with a quick paper and pencil test, and call that a measure of intelligence.Then, based on those results, they grade people on a linear scale from moron to five-star genius, while ignoring that the moron may have the musical genius of Derek Paravicini, or the visual and artistic gifts of Stephen Wiltshire, whereas the five-star genius may be incapable of carrying a tune, recognizing a face, or cutting a two by four square.

The IQists justify this presumption by claiming that their tests measure a unitary property of the mind, which justifies their fascistic claim to be able to grade all of humanity on a uni-dimensional scale, thereby establishing a global pecking order for once and all.

But the evidence, neurological, anatomical and even, psychometric, refutes their claim. Intelligence is not a unitary property of mind and, therefore, IQ is a misnomer for whatever it is that the IQist’s measure. Moreover the use of the term IQ involves an egregious abuse of millions of people subject to the IQist measure, since it labels them in a way that in no way reflects the multiplicity of their natural talents.

As RR88 points out, the IQist’s notion of intelligence has no “construct validity,” nor therefore, any operational definition. Instead, reliance is placed on a circular definition. IQ tests measure intelligence, therefore, intelligence is what IQ tests measure.

If Psychometricians would abandon their arrogant pretensions, and acknowledge that what their tests attempt to measure are specific abilities, these the product of both genes and environment, they would be in a position to offer a modestly useful service with minimal risk of doing harm. Unfortunately, the pursuit of power has always been the besetting sin of the students of mind, from Freud to Watson, to Skinner and Eysenck. In view of the absurdity of their past blunders, one might have hoped for greater humility in today’s practitioners, but it seems one will hope in vain.

…. which are NIET IQ…

Sanlonguito de mi corazón!!

… to be a mathematical genius or a physics genius or any other kind of genius you may need some basic aptitude measurable as IQ. But beyond a basic requirement for IQ, rarely more than 120 or so, genius, or intelligence as that term is understood by the vast majority of users of the English language, is a matter not of IQ, but of temperament, education, environment, and the mere chance of being born in the culture, in the age, and in the social class in which a particular form of genius can emerge. (emphasis added)

AS ALWAYS happen ”intelligence-researchers” barely study creativity… so, they to be drifting.

In order of importance or relevance:

- Temperament/IQ or cognition;

- environment

– Education is part of temperament {motivation + appropriate environment] but more about motivation because even in not ideal environment you can create things, i mean, in adverse situations.

Iqists

hate;
don’t understand;
don’t care about creativity, they are full of biases about anything which are IQ. I don’t think it’s sound too much strawman.

Any position in DNA that has a non-zero variance is from definition SNP.

Do you happen to recall what that S in SNP stands for? It does not work as you said. A good example is indels (insertions and deletions, but you should know this). They can add or remove amino acids completely, OR if they are a non-multiple of three nucleotides in length they can render a coding sequence into complete gibberish. The downstream SNPs might still exist (SNP detection is a local thing), but they are meaningless.

If there were no SNP’s we all would be clones of one master original.

Nice attempt at rhetoric. Do you think statements like that constitute an argument?

If you have 0.01% variant how much variance you expect it possibly can explain if it is present only in 1 of 10,000 subjects?

Not a great deal individually. A worthwhile point. Far from decisive though.

Those rare variants can be important. IIRC I recently linked a paper here looking at rare variants in disease which found some quite large effect sizes for some of them.

And then there is the question of how many strongly deleterious rare variants there are. If one explains on the order of 0.001% of variance and there are tens of thousands of them that becomes important.

One can make CNV to go away conceptually by changing the mapping, redefining the address book of SNPs and then it all comes down to SNPs again.

I’m not sure how well SNP chips can detect absolute numbers of CNVs. There is also the issue of mutations occurring in the same/different copy/ies of a gene. This is similar to the issue of whether mutations occur on the same DNA strand. I don’t believe current SNP chips detect this (it’s called phasing if you want to delve deeper) and it may very well be another source of “non-SNP” heritability. In short, the issue is that two disabling mutations have a larger phenotypic effect if they occur on different DNA strands than if they occur on the same strand–once a gene is broken (well, I am simplifying, there are degrees) it is broken. Breaking both copies is generally worse than breaking one copy twice. To be clear, this is not just nonlinearity. It also requires knowing which strand each SNP is on (and then quantifying the combined effects, which is hard!).

I think there is also an issue with whether or not CNVs capture entire functional genes. CNVs are not just sums of SNPs. The connectivity matters.

Utu, I know I come off as patronizing when I say this, but you really do need to take a genetics class if you want to engage on this topic to the depth you attempt.

P.S. To add something a bit more interesting to this comment, this paper and its abstract give some idea of how common variation beyond the “typical” SNPs is: http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0059494

Abstract

Whole genome sequencing studies are essential to obtain a comprehensive understanding of the vast pattern of human genomic variations. Here we report the results of a high-coverage whole genome sequencing study for 44 unrelated healthy Caucasian adults, each sequenced to over 50-fold coverage (averaging 65.8×). We identified approximately 11 million single nucleotide polymorphisms (SNPs), 2.8 million short insertions and deletions, and over 500,000 block substitutions. We showed that, although previous studies, including the 1000 Genomes Project Phase 1 study, have catalogued the vast majority of common SNPs, many of the low-frequency and rare variants remain undiscovered. For instance, approximately 1.4 million SNPs and 1.3 million short indels that we found were novel to both the dbSNP and the 1000 Genomes Project Phase 1 data sets, and the majority of which (∼96%) have a minor allele frequency less than 5%. On average, each individual genome carried ∼3.3 million SNPs and ∼492,000 indels/block substitutions, including approximately 179 variants that were predicted to cause loss of function of the gene products. Moreover, each individual genome carried an average of 44 such loss-of-function variants in a homozygous state, which would completely “knock out” the corresponding genes. Across all the 44 genomes, a total of 182 genes were “knocked-out” in at least one individual genome, among which 46 genes were “knocked out” in over 30% of our samples, suggesting that a number of genes are commonly “knocked-out” in general populations. Gene ontology analysis suggested that these commonly “knocked-out” genes are enriched in biological process related to antigen processing and immune response. Our results contribute towards a comprehensive characterization of human genomic variation, especially for less-common and rare variants, and provide an invaluable resource for future genetic studies of human variation and diseases.

You mention IQ 120. My point was that you were probably wrong. Nothing special about IQ 120.

I did not say or suggest that there was anything special about IQ 120. What I said, and what I see no reason to abandon was:

… to be a mathematical genius or a physics genius or any other kind of genius you may need some basic aptitude measurable as IQ. But beyond a basic requirement for IQ, rarely more than 120 or so, genius, or intelligence as that term is understood by the vast majority of users of the English language, is a matter not of IQ, but of temperament, education, environment, and the mere chance of being born in the culture, in the age, and in the social class in which a particular form of genius can emerge. (emphasis added)

What the actual IQ cutoff is for mastery of the higher math, I don’t know, but if it is possible for someone with an IQ of, say, 95, or 105 or even 115 to obtain a first class degree in mathematics, then what possible value can there be in measuring IQ?

What I challenge is the concept underlying the measurement of IQ; namely, that it represents a unitary feature of mind. This is unlikely for many reasons, not least, the modularity of the brain. Moreover, in some circumstances, there is direct evidence of cognitive capacities with unique variances. for example:

Delis, DL, et al. 2003. The myth of testing construct validity using factor analysis of correlations with normal or mixed clinical populations: Lessons from memory assessment (J Int Neuropsychol Soc. 2003 Sep;9(6):936-46.)

…cognitive measures that share variance in the intact brain-thereby giving the facade of assessing a unitary construct-can dissociate and contribute to unique variance in the damaged brain…

Education in these genetic studies almost always refers to years of education. That is because years of education is collected along with other key personal information in cross-sectional DNA databases. It is simply a data availability issue, and obviously an important limitation, that this variable proxies for “education” understood more broadly.

Based on… quantitative studies…

What is wisdom for you* I thought there is a way to ”measure” this.