« Nerds and value creation | Main | Roissy’s blog is gone »

July 28, 2011

Comments

Why do you use the GSS for studying correlations with intelligence, when there are data sets like the NLSY (used by Murray & Herrnstein in "The Bell Curve) that are freely available and have a much better IQ test, larger and more representative samples sizes, etc. than the GSS?

[HS: The GSS is the simplest and most easily accessible dataset, and the SDA interface creates nice HTML charts that can be pasted into blog posts.

NLSY is complicated and requires you to use SPSS, and I'd have to buy a copy (not cheap) and learn how to use it.]

JL,
Why shouldnt we have an extra data set?

You don't need to buy SPSS to analyze data from the NLSY. You can download raw data and use free software such as R.

Of course it's much easier to use the GSS, but I think many of your findings are spurious due to unrepresentative samples and unreliable variables.

You can just torrent SPSS.

Off topic but anyone else notice Roissy's blog is gone? Hopefully it's just a temporary hiccup but if not then today is a dark day for all men.

Yes its been known for some time that people who do best in the maths and sciences actually have the highest verbal scores but choose STEM out of preference, because verbal scores correlate best with general intelligence.

Asians seem to be an exception - they score low on verbal yet are very STEM, our of preference, it would seem. Asians do best on visuo-spatial and on the Raven Matrices (a controversial test - some consider it the best measure of g, but it is also the only IQ subset that has experienced the Flynn Effect to a massive degree, leading many to conclude that it does not measure intelligence at all, because the alternative explanation, that the RM is an accurate measure of g and we are all geniuses now compared to our ancestors of 50 years ago is obviously absurd, not to mention the fact that in the same time period other things highly correlated with IQ, like SAT tests, have declined, which would lead to the absurd conclusion that something in the past 50 years elevated intelligence dramatically while at the same time decreasing something that is highly correlated to intelligence.)

The idea that STEM does not require verbal is caused by seeing so many Asians in STEM, who have low verbal, but this is misleading, as cultural preferences account for Asian STEM concentration.

STEM takers probably include people who immigrated here or children of immigrants, who may not have spoken English at home. I wonder how different scores would be if you excluded these people...

I heard some Chinese students on radio today. They were American studies postgraduates who had never lived outside china. I was struck by their impeccable English. They were grammatically faultless and their vocabulary was adequate and their choice of words unimpeachable. This leads me to think that even the non-STEM-stereotype Asians will be a formidable force academically.

Back on topic: high-end STEM people are high in G. I have always acknowledged this, despite being a verbal freak (99 pb) and a quant mediocrity (75 pb). I like to think that specialist wordsmiths have some magic spark, but I have never met an eye surgeon who is incapable of stringing a few words together.

Interesting to see all the rationales for why STEM would have lower scores than English majors. I think it's less an overabundance of Asians than that English majors have effectively been taught the test. Even a complete idiot who's been taught vocabulary for four years should have a pretty high Wordsum score. Wordsum does correlate somewhat with intelligence, but that doesn't mean there are no distortions in the data.

"...things highly correlated with IQ, like SAT tests..."

This is a misunderstanding of what IQ is and what IQ tests are.

The SAT is not "highly correlated with IQ", the SAT is an IQ test.

"The SAT is not "highly correlated with IQ", the SAT is an IQ test."

A comment like that ought to reinvigorate this thread if anything will! I don't disagree, I just don't know enough of the psychometric literature to say. But, it's widespread use would seem a good indicator.

Off topic, but has anyone ever solved a Rubix cube from scratch, or know anyone who has? I mean no algorithms, hints, etc. I'm very curious as to their other intellectual stats/achievements. I have given it a solid hour or more on multiple occasions and can at best manage one face plus one to two rows around it. I wonder what the floor is for visuospatial or other abilities.

The SAT *tests* far too much knowledge to be a pure IQ tests. It obviously measures knowledge+intelligence. Therefore it only correlates to IQ.

To be fair, IQ tests test for knowledge to some extent as well, as you need to know words to do well on the verbal subtests and you need to know math to do well on the math subtests, but far less than the SAT.

The Raven Matrices is one of the few IQ subtests that dont involve any knowledge at all and for that reason some people consider it to be the best measure of general intelligence. However the RM is a very controversial test as I explained above and some people dont consider it a measure of general intelligence at all.

J,

You don't know what you're talking about.

Nicolai,

I stand corrected. I defer to your obviously superior knowledge of this subject.

Non-verbal tests also test knowledge.

Those Wechsler subtests least affected by the Flynn effect are vocab, info, and arithmetic.

Explain that.

The achievement/intelligence test distinction is impossible to make in practice.

The SAT and other such tests correlate as well with self-described IQ tests as those self-described IQ tests correlate with each other. They cannot be distinguished from the psychometric point of view, and poorly educated people with high IQs don't exist.

"Non-verbal tests also test knowledge."

Not really. Most of them measure fluid reasoning, as opposed to acquired crystalized knowledge.

"Those Wechsler subtests least affected by the Flynn effect are vocab, info, and arithmetic."

That's because these tests are culturally biased against the younger generation (knowledge is always changing) so they fail to fully reflect their rise in intelligence as well as fluid tests do. Also, studies of babies born malnourished show they are least impaired on these types of subtests. Since the flynn effect is caused by nutrition, it makes sense that these tests would show the smallest gains.

"The SAT and other such tests correlate as well with self-described IQ tests as those self-described IQ tests correlate with each other. They cannot be distinguished from the psychometric point of view, and poorly educated people with high IQs don't exist."

The SAT only correlates with culture reduced IQ tests among people with similar schooling and culture. If you gave the SAT to a representative sample of the entire world, it would not correlate well at all with real IQ tests like the raven which don't assume much cultural knowledge.

English Lit. isn't easy. It just isn't that useful. The fact that an ostensibly intelligent persons who don't have family to support them, marry them off, or get them a job, still major in it makes one question their choice.

"AM in NAM", as an indian, you're clearly not a NAM. Unlike you, I'm a real NAM, not only that, a member of the worst class of NAMs (you know who I'm referring to). And unlike you, I can't say I make 6 figures but I do well for myself. Oddly enough, I've never been, as you put it, "racially abused" in the 9 years I've been here, definitely never been called a chimp (at least not to my face). The only racism I've seen is on the web, by whites emboldened by the (perceived) anonymity that the internet offers. That being said, HBD-ers are the most rational in their racism, because it's based on facts and a mountain of evidence that is difficult to ignore. But a part of me still loathes HBD-ers because they typically feel more hatred than pity for races whose only crime is that they got the short end of the stick in the evolution lottery.

Of course there are few (like Steve Sailer) whose view on NAMs (blacks especially) tends towards the "pity" end of the spectrum; Sailer seems to look down upon blacks with a mixture of amusement and a little bit of pity, and at least he seems to be in awe of the one good quality we possess (athleticism). I've lost count of how many posts he's done on black athleticism.

"Not really. Most of them measure fluid reasoning, as opposed to acquired crystalized knowledge."

That's why these tests ARE affected by the Flynn Effect, and tests of verbal knowledge aren't?

50 years ago the rural ill-educated Afrikaners scored like black Americans compared to the urban well-educated British South Africans. And their lower scores were on the non-verbal subtests.

It's pointless arguing about IQ with people who have a low IQ.


"The SAT only correlates with culture reduced IQ tests among people with similar schooling and culture. If you gave the SAT to a representative sample of the entire world..."

You've just been hoist by your own petard Whydont.

The "culture-fair" Ravens has had a 20 pt increase in 50 years for the Dutch.

And "culture-fair" tests must be non-verbal, but tell me what makes a man a man. Birds fly, fish swim, men talk. A non-verbal intelligence test is going to miss what is essential to being human.

There are no "culture-fair" tests. The idea that there are requires the assumption that all knowledge is verbal, that culture makes no difference in non-verbal skills.

The bottom line is that the "experts" in intelligence testing are PhDs in psychology and therefore they are DUMB, STUPID, etc.

"That's why these tests ARE affected by the Flynn Effect, and tests of verbal knowledge aren't?"

Exactly! The verbal knowledge tests are not sensitive and culture neutral enough to pick up subtle nutrition based improvements in neurological development that have been occurring over the last one hundred years or so. Just as nutrition has made people taller, brain size and complexity have also been increasing and it's been showing up on the best most culture fair measures of fluid reasoning.

"50 years ago the rural ill-educated Afrikaners scored like black Americans compared to the urban well-educated British South Africans. And their lower scores were on the non-verbal subtests."

Blacks reared in Africa average IQ 67, but because of better nutrition full-blooded blacks raised in America average IQ 80 and several inches taller. Indians in India average IQ 81 and very short height, but Indians in America are normal height and high achievers. You take any race and improve their nutrition starting before birth and they will grow up much taller and smarter than they would have been, however genetic differences prevent races raised in the same time and place from being equal. Similarly, nutrition has made both men and women taller, but the genetic height gap between them remains constant

Those who explain the Flynn Effect with nutrition, which does not include James Flynn himself, are zoophiles.

http://weirdnews.about.com/b/2011/07/10/zoophilia-death-woman-killed-in-dog-sex-incident.htm

"The SAT only correlates with culture reduced IQ tests among people with similar schooling and culture. If you gave the SAT to a representative sample of the entire world, it would not correlate well at all with real IQ tests like the raven which don't assume much cultural knowledge."

You've just been hoist by your own petard Whydont.

1. The Raven's DOES assume cultural knowledge. It just doesn't assume verbal cultural knowledge. In 50 years IT has been affected by the Flynn Effect by 20 POINTS among the Dutch.

2. Fish swim, birds fly, men talk. What sort of IQ test doesn't test ability at what is MOST essentially human?

"...and a mountain of evidence that is difficult to ignore..."

I can ignore almost all of it. Black Americans are untouchables like Koreans in Japan and the Red Legs of Barbados.

I know this because I've seen whites talking and acting black without affectation. Apparently they had grown up around blacks.

Race is irrelevant. If you want a better world then you want smart people to have more children and stupid people fewer. Not all blacks are stupid, and a lot of whites are.

http://www.youtube.com/watch?v=N8QEIaATPis

The interesting thing is that what Julian Bond says here has been investigated. Accoding to Nisbett, IIRC, a group of very high IQ black children in Chicago were found to be indistinguishable from other blacks in lightness/darkness.

Of course the HBD-er might say that only stupid whites deign breed with blacks.

Nicolai, how does the Raven Matrices test knowledge.

The way I explain that there is no Flynn effect on the vocab and math subtests is by saying that those tests where there IS a Flynn effect - such as the Raven Matrices - dont really measure intelligence, whereas the vocab and math tests do test intelligence. Thats the obvious conclusion. There is incredible resistance in the psychometric community to admitting this fact because the RM is the most culturally reduced test, and psychometricians really WANT to believe in its efficacy.

Youre right, a complete separation between knowledge and achievement is not possible. But its all a question of degree.

A history test is not an IQ test, but a vocab reasoning tests is, even though both test knowledge to some defree. The question is one of degree. The whole POINT of IQ was to try and reduce the knowledge content as much as possible and isolate reasoning power. To fudge this crucial distinction now is disingenuos.

High IQ people with poor educations have and continue to exist - lots of people know no algebra or geometry and will do terrible on the math SAT but excellent on IQ tests, even the math subtests, where the knowledge threshold is much lower

Just because something corelates with IQ scores as well as IQ subtest corelate with each other does not mean it IS an IQ test. All sorts of things correlate well with IQ scores - if I give some a quesionnaire (test) about how much crime he commited and scored it, the score would correlate well with his IQ score. Doesnt make a crime questionnaire an IQ test.

Nicolai, the whole point of the IQ test was to isolate reasoning power from mere measures of knowledge. To that end, the knowledge content is greatly reduced in IQ tests. It cannot be separated entirely, no, but I never claimed it could.

A criminal history test probably correlates well with IQ scores, but a criminal history test is not an IQ test. To say then that from the psychometric point of view there is no difference between a criminal history test and an IQ test is obviously absurd. Lots of things that can be measured by tests and scored correlate well with IQ but are not IQ tests.

As for the Flynn Effect, the obvious conclusion is that the tests on which it appear dont really measure intelligence.

Uneducated smart people undoubtedly exist - people with no interest in math might do terribly at the math SAT because they dont know algebra, but do well on the math IQ, because the knowledge threshold is much lower.

All about knowledge threshold, my friend.

"1. The Raven's DOES assume cultural knowledge. It just doesn't assume verbal cultural knowledge. In 50 years IT has been affected by the Flynn Effect by 20 POINTS among the Dutch."

The Dutch have also grown several inches taller over the past 50 years, so by your logic, measuring tapes must measure cultural knowledge too.

"2. Fish swim, birds fly, men talk. What sort of IQ test doesn't test ability at what is MOST essentially human?"

Talking requires abstract reasoning and that's what the raven measures.

"As for the Flynn Effect, the obvious conclusion is that the tests on which it appear dont really measure intelligence."

No the obvious conclusion is that certain parts of intelligence can be improved by nutrition, so as society grows taller with bigger, better nourished brains, this translates into superior performance on the purest measures of IQ.

"High IQ people with poor educations have and continue to exist - lots of people know no algebra or geometry and will do terrible on the math SAT but excellent on IQ tests, even the math subtests, where the knowledge threshold is much lower"

Smart people YES. High IQ people NO.

"Just because something corelates with IQ scores as well as IQ subtest corelate with each other does not mean it IS an IQ test."

1. The other correlates you mention do not correlate as well as IQ tests correlate with one another.

2. What is an IQ test? Theoretically, from the point of view of psychometrics, it is a short way to estimate how well the test taker will do on ALL/ANY "cognitive" tests. Such an estimate is possible, because all such tests correlate with one another. So again any test which correlates with self-described IQ tests as well as these correlate with each another is BY DEFINITION an IQ test.

3. The Raven's is an outlier among IQ tests, because it is not a battery. It is homogeneous. It's correlation with other IQ tests is not very good. Supposedly it is a good measure of the ridiculous g, though.

If Einstein had been raised on a cattle ranch in Namibia he wouldn't have been Einstein and he would not have even been "an Einstein".

If you don't like that then have him raised by wolves.

Whatever.

It is impossible to distinguish native ability from acquired ability.

Because someone makes a test he calls an IQ test and "tries" to make it just as accessible for the wolf-child as for the trust fund kid is irrelevant to whether or not it is an IQ test, and such a person is sellling snake-oil or he's just stupid.

Whether or not the Raven's tests knowledge is a matter of sematics. The point is that in the last 50 years the ability to do the Raven's has improved as a result of changes in the intellectual environment, perhaps video games or television. This is a REAL increase in intelligence and a REAL increase in g. That such a real increae may be effected by a change in environment is anathema to idiots like Jensen but shouldn't be.

"Talking requires abstract reasoning and that's what the raven measures."

That's totally retarded.

"No the obvious conclusion is that certain parts of intelligence can be improved by nutrition, so as society grows taller with bigger, better nourished brains, this translates into superior performance on the purest measures of IQ."

What an idiot. That such utter bullshit is propounded by some IQ "experts" only shows how low the bar is for a PhD in psychology.


Nicolai, dude, a criminal history test correlates better with IQ score than many of the IQ subtests correlate with each other. Nevertheless a criminal history test is NOT an IQ test. I simply dont understand how you cannot get this. The point is crystal clear. Its almost laughable that Im arguing this.

Tests of other things that dont fit the definition of a cognitive ability, but nevertheless correlate well with tests of cognitive ability, are still not tests of cognitive ability. A personality tests is not a test of cognitive ability.

The comments to this entry are closed.