Sunday, August 28, 2005 ... /////

BBC: sensitive IQ tests

The attitude presented in the following paragraph is a joke, of course.

BBC has apparently abused the freedom of speech in England. They were not afraid and ashamed to publish the outrageous results of IQ tests designed to measure "general cognitive (spatial and verbal) ability" that two British researchers (Paul Irwing and Richard Lynn in British Journal of Psychology) gave to 80,000 people plus 20,000 extra students; well they summarized some previous tests but it does not really matter. What a crummy article.

• Below 14 years: no measurable difference between boys and girls
• Above 14 years: IQs differ by 5 points
• IQ 125 and above: 2 vs. 1 ratio
• IQ 155 and above: 5.5 vs. 1 ratio

Their story immediately became the most discussed news story in the blogosphere. To balance the viewpoint, an alternative research with the opposite result was performed by Barry Sheerman, a British MP: women are cleverer than men, he said.

The word "research" used for Sheerman's statements was another joke, of course.

Now: most IQ tests are not terribly deep (and one can train herself or himself to get better results) - and one can't really deduce anything certain from one silly number. Nevertheless, if a question should be studied rationally, something like that had to be done anyway and the correlations between the resulting IQ and other things will exist. As far as I know, they (mostly?) used the Raven Standard Progressive Matrices which is respected as a good test of the pure intelligence factor g.

Earlier in the summer, another extremely bright and mysterious sinner named La Griffe du Lion (equivalently, the Zorro of statisticians) has looked at the power of Gaussians and the different second moments or variances. He showed how Summers' speech on January 14th could have looked like and one of his or her conclusions is that a female Fields medal is expected once per 103 years.

Disclaimer: Of course, your humble correspondent respects the official opinion of Harvard University - as codified by the Congress of Women's Committees and the FAS faculty meeting on 3/15 - which says that two female Fields medals are expected every four years; one prize per 2 years. No doubt, the fact that du Lion's value seems closer to reality is a measurement error or a result of discrimination. The Reference Frame is not responsible for the content of external internet sites such as BBC.

The previous paragraph was a joke.

More seriously, this research may be interesting, it can be the best approach that a purely sociological and psychological research can take, and it probably implies something but I still prefer neurobiology that tries to figure out the true reasons behind all of this.

Even more seriously, The Reference Frame disagrees with some political statements by Richard Lynn. Amusingly, I am probably the first person to conjecture that La Griffe du Lion and Richard Lynn is the same person. Of course, du Lion could also be Charles Murray whose new and up-to-date extensive analysis of group differences is officially released next month. Or anyone else among the known names, including one of our friends from Harvard. But my guess is Lynn anyway. Feedback welcome. ;-)

Richard Lynn and Paul Irwing became very unlikely scientific bedfellows because Paul Irwing is kind of left-wing, politically correct person who would have been "happier if this fact were never found out or if the result were different". It just happened that someone had convinced him that suppressing a certain kind of data was scientifically dishonest, and Irwing agreed.

One of the things that fascinates me is that most people apparently believe that all women have to dismiss the research immediately as permanently unacceptable. I beg to disagree and many of my female friends are counterexamples. Gutsy women like Jennifer Warwick know that time will tell whether Lynn's research is good science; and it is definitely an opportunity to test and explore our own biases and prejudices.

snail feedback (15) :

Dear Lumos,

'Now: most IQ tests are not terribly deep (and one can train herself or himself to get better results) - and one can't really deduce anything from one silly number.'

I thought you had the world's second highest IQ, so it is very humble of you to dismiss IQ like this. It makes me feel really bad for my one or two negative posts I added about your defense of ST on Peter Woit's blog.

BTW, t'Hooft has suggested that the holographic conjecture (based on Bekenstein bound) implies that a 5-D spacetime can be viewed as 4-D spacetime with the fabric of space as the fifth dimension. Bekenstein says ( http://eskesthai.blogspot.com/2004/11/information-in-holographic-universe.html ):

'Superstring theory rules in the 5-D spacetime, but a so-called conformal field theory of point particles operates on the 4-D hologram. A black hole in the 5-D spacetime is equivalent to hot radiation on the hologram--for example, the hole and the radiation have the same entropy even though the physical origin of the entropy is completely different for each case.'

Can the spacetime fabric as the fifth dimension be treated by causality? Can the mechanism of gravity be extended from LeSage gravity (see 5 star book by Edwards on LeSage: http://www.amazon.com/exec/obidos/tg/detail/-/0968368972/103-7201186-4953411?v=glance ).

Since Peter Woit won't allow me to post anything positive on his blog, I've put my views here: http://eskesthai.blogspot.com/2005/07/history-of-gravity-and-equivalence.html

Yours sincerely,
Nigel

Lubos:

Your today's post reminded me of an event when you and I actually spoke together.

Place: The String Palace(Rutgers), public computer room

Year: Around 2000

Time: Early morning

Who: You, some other person, and I (Slawomir Piatek)

You and I used to be a lot in that room, but for some reason, we never spoke, except for one time when you found some IQ-test-related problem on the Internet involving a triangle... some kind of illusion. You were very amused by this problem and started a conversation in the room. Then, we went back to our things: you to either banging on a keyboard or to dealings with Ada Banks, and I to staring at stars on a computer screen.

Slawomir Piatek

Dear Slawomir,

we should have talked more often! :-) Was it the triangle at physics.rutgers.edu/~motl/trojuhelnik.gif? I kind of remember your story but not sure about your face.

Was there Tolik Morosov in the room?

All the best
Lubos

Lubos:

I do not recall the other person very well... She might have been that undergraduate student who worked with Jack Hughes (Astronomy). She also spent many many hours in that room.

The "triangle problem" had something to do with Pythagoras' theorem... It did not look like the one you posted. Ah, how memory tends to fade away...

Slawomir Piatek

All I can say is that Richard Lynn is one who has made the mistake asserting that

Average( f(X)) = f( Average(X))

Here X is a set of scores on Raven's Progressive Matrices Test which have a highly non-linear relationship to IQ; he converts an average Raven's test score to an average IQ based on the above.

Kamin on Lynn:

To illustrate what Lynn has done, consider a small thought-experiment.  We travel to Africa and give the Matrices test to a large number of children, all aged 13.5.  Half the children have raw scores of only 13 correct answers, because they did not get the point and are merely guessing on the multiple choice test.  The other half do get the point and all have raw scores of 56.  The British standardization of 1979 indicates that those two raw scores fall at the 1st and the 99th percentiles, respectively.

Thus the average percentile score of the children is 50, corresponding to the exact center of the bell curve. The center of the bell curve of course implies an average IQ of 100.

But Lynn would seize upon the fact that the average raw score was 34.5.  That score corresponds to the 8th percentile in the standardization sample.  Lynn, consulting the bell curve, would observe that the 8th percentile of a normal distribution corresponds to an IQ of 79, and would report that figure as the average Negroid IQ. Murray and Hernnstein would believe him; he is, after all, their expert.

You want to take him seriously, Dr. Motl, please do, it will certainly add to your own credibility. Or perhaps you will dig through the literature, and find out that even a Harvard professor such as Hernnstein can be Not Even Wrong.

All I can say is that Richard Lynn is one who has made the mistake asserting that

Average( f(X)) = f( Average(X))

Here X is a set of scores on Raven's Progressive Matrices Test which have a highly non-linear relationship to IQ; he converts an average Raven's test score to an average IQ based on the above.

Kamin on Lynn:

To illustrate what Lynn has done, consider a small thought-experiment.  We travel to Africa and give the Matrices test to a large number of children, all aged 13.5.  Half the children have raw scores of only 13 correct answers, because they did not get the point and are merely guessing on the multiple choice test.  The other half do get the point and all have raw scores of 56.  The British standardization of 1979 indicates that those two raw scores fall at the 1st and the 99th percentiles, respectively.

Thus the average percentile score of the children is 50, corresponding to the exact center of the bell curve. The center of the bell curve of course implies an average IQ of 100.

But Lynn would seize upon the fact that the average raw score was 34.5.  That score corresponds to the 8th percentile in the standardization sample.  Lynn, consulting the bell curve, would observe that the 8th percentile of a normal distribution corresponds to an IQ of 79, and would report that figure as the average Negroid IQ. Murray and Hernnstein would believe him; he is, after all, their expert.

You want to take him seriously, Dr. Motl, please do, it will certainly add to your own credibility. Or perhaps you will dig through the literature, and find out that even a Harvard professor such as Hernnstein can be Not Even Wrong.

Dear Arun,

if this is really everything you can say, let me admit that it does not seem as terribly much to me. ;-)

Moreover, f(average(x))=average(f(x)) is true at linearized order for f(x), and if this error has ever been done once, be sure that it is not being made now.

Look at du Lion's page where, in fact, your inequality (especially in the context of the non-linear relation between GDP and IQ) plays a relatively important role at some places.

These guys are extremely bright, and the attempts to create the illusion that they're not simply can't work, Arun. Forget about it.

Best
Lubos

Or let us take another example, in apartheid South Africa, naturally, white students at school were found to have higher test scores on various tests than blacks. Except Crawford-Nutt, in 1976, found that giving clearer instructions to his test subjects (black students in Soweto, South Africa) resulted in them scoring at the white norm on Raven's Progressive Matrices. Lynn ignored this study as a flawed one in his compilation of African IQ. ( One might ask, why did blacks need instructions; one would find out the differences in conditions in schools, that for whites, the tests were typically conducted by the school psychologist.)

On the other hand, Lynn is perfectly happy to take published scores from papers with caveats in the paper like "the knowledge of English of the majority of black testees was so poor that certain [of the] tests...proved to be virtually unusable", and happily convert these into IQs.

One last thing, anticipating an argument, anticipating because I'm less and less inclined to return here - Lynn could be wrong in all his previous research and right in the one cited here.

It is certainly possible. On the other hand, having seen enough of the character, I would not trust him to be honest in what he reports. The mistakes Lynn makes are not of the honest mistakes scientists make, they are "active" errors, i.e., one has to go out of the way to fall into those errors.

Dear Arun,

I certainly agree with you that if someone gives "clearer instructions" how to fill out a test, it is reasonable to expect that the particular group (in this case, in Soweto) will have a better score in average. Is not it what happened? ;-)

Also, I think that if one tries to compare some people or groups in a test, all of them must be given the same tests, instead of giving "clearer instructions" to a subgroup. If this rule is not satisfied, the particular research is flawed. Do I really understand well that you disagree with this trivial comment?

Let me also mention that the new work has two authors. The other one is a politically correct leftwinger named Paul Irwing who would prefer a different result but who agrees that doubts no longer exist. Your ad hominem comments against Lynn - apparently used as a method to argue against their analysis - are irrational and kind of painful.

All the best
Lubos

Arun, unless you have already left the blog, before you would throw up ;-), another comment.

Incidentally, if you search for the word "grotesque" in Murray's new article, you will see how the analyses of these statements about some subtle reasons like the "stereotype threat" (or "unclear instructions") are misinterpreted in the media.

The relationship between the Triangle problem and pythagoras theorem is about a series of proofs of Pythagoras which were done by cutting and pasting triangles (instead Euclides proof via triangles of equal basis and height). Probably that webpage contained also some of these proofs, or Lubos was aware of them from other source (very popular between physicists, it was). The triangle problem can be exhibited when discusing about the rigour, or lack of it, of visual proofs.

(the visual proof in this case shows that 3/8 = 2/5, which is potentialy important because as we know 3/8 is the sine of Weinberg angle. Er, just joking :-D Or self-parodying )

Leucipo, it seems as a good point. Nevertheless, your 2/5 "=" 3/8 "=" 5/13 is about the triangle problem I posted: these invalid identities lead one to believe that the tilted line is actually straight. In reality, it's broken in two different ways, and the difference of the areas make up one square. What's its relation with the Pythagorian theorem?