g, a Statistical Myth

[Ahknaton]

http://www.cscs.umich.edu/~crshalizi/weblog/523.html

[President Barbicane]

Very interesting. I had read Cosma Shalizi's previous writings about the heritability of IQ, and I was not convinced. This work, however, is better. I now believe that there may be several different, unrelated traits that together explain people's performance on IQ tests.

[Macrobius]

LOL 'principle component analysis'.

[KevinDeBurgh]

http://www.newscientist.com/article/mg20427321.000-clever-fools-why-a-high-iq-doesnt-mean-youre-smart.html?full=true


*article is copyrighted so I'm not pasting it here.

[Angler]

I'll read that article later on, but for now I want to say this. I'm quite convinced that cognitive tests are generally useful in spite of their flaws. Measuring intelligence will never be as precise as measuring height or weight, but tests for the purpose tend to provide pretty decent estimates. They are particularly useful for comparing large populations within a single society, e.g., blacks and whites of equal socioeconomic status who were born and raised in America.

Still, I don't believe that intelligence is unitary, and there is considerable evidence for the modular nature of mental abilities. The evidence includes people who are prodigies in fields like math or chess but aren't unusually bright in other areas; stroke and brain damage victims who lose certain specific kinds of abilities (e.g., linguistic) but whose other mental abilities remain intact; data from brain imaging studies; and perhaps more.

Why, then, do measurements of performance on different kinds of mental tasks tend to be positively correlated? I think it's simply because people whose brains are physically well-developed and high-functioning in certain regions tend to have similarly good development and functioning in other regions of the brain. To the extent that, say, ability to understand grammar is correlated with the ability to learn geometry, the basis of the correlation might be nothing more than common genetic and environmental factors that influence development in disparate sub-volumes of the brain.

Consider a physical analogy. Obviously muscular strength is not unitary, since muscles exist separately from one another. The biceps in my arms are separate from the hamstrings in my legs. Yet I think it's clear that a person who tends to be thinly-muscled in some areas of his body also tends to be scrawny elsewhere. You don't see many people who have enormous thighs but 12-inch upper arms; likewise, few people have 20-inch upper arms but chicken legs. That's not to say you don't see uneven development in some people, especially if they're on roids; but the point is that there's a clear correlation between strength in one part of a person's body and his strength elsewhere. The same is true of fatness or thinness: how many fat people have big, doughy arms but "ripped" thighs? It might happen, but not very often.

Interestingly, I've read that the higher one's "overall intelligence" appears to be, the less his mental abilities appear to be positively correlated. In other words, those who tend to score around 90-110 on the Stanford-Binet might have verbal, math, and spatial abilities that are closely correlated, but those who tend to score 150 are more likely to have brains that are more "specialized." This observation might shed some light on the question of what g really is. However, even the adult brain is capable of adapting to the demands put on it, and that could be a confounding factor: how do we know if high-IQ people were born with more specialized brains or, alternatively, if their brains became that way due to greater adaptability coupled with unique efforts directed at particular kinds of mental tasks?

[KevinDeBurgh]

I don't necessarily have a strong opinion either way yet since I've never seriously researched this before. I just posted that article because it seems apropos here. I don't necessarily support everything or anything said in articles I post here on the phora but of course most articles I post or link to but I probably tend to agree with for the most part.

[Angler]

I don't necessarily have a strong opinion either way yet since I've never seriously researched this before. I just posted that article because it seems apropos here. I don't necessarily support everything or anything said in articles I post here on the phora but of course most articles I post I probably tend to agree with for the most part.
Actually, I was referring to Ahknaton's article in my post. :) But I'm going to take a look at yours, too.

[japanese anime]

http://www.cscs.umich.edu/~crshalizi/weblog/523.html
Do you expect someone who calls it "principle component analysis" instead of "principal component analysis" to be well versed in psychometric literature?
This person is barely qualified to clean my shoes, never mind write an essay on g.

[Ahknaton]

Do you expect someone who calls it "principle component analysis" instead of "principal component analysis" to be well versed in psychometric literature?
Yes.
This person is barely qualified to clean my shoes, never mind write an essay on g.

The author:

http://en.wikipedia.org/wiki/Cosma_Shalizi

Cosma Rohilla Shalizi (born February 28, 1974) is an assistant professor in the Department of Statistics at Carnegie Mellon University in Pittsburgh.

Shalizi is co-author of the CSSR algorithm, which exploits entropy properties to efficiently extract Markov Models from time-series data without assuming a parametric form for the model.[1]

Shalizi writes a popular science blog "Three-Toed Sloth".

In 1990 he was accepted as a Chancellor's Scholar at the University of California, Berkeley, and completed a bachelor's degree in Physics. Subsequently, he attended the University of Wisconsin–Madison where he received a doctorate in physics in May 2001. From 1998 to 2002, he worked at the Santa Fe Institute, in the Evolving Cellular Automata Project and the Computation, Dynamics and Inference group. Afterwards, from 2002 to 2005, he worked at the Center for the Study of Complex Systems at the University of Michigan in Ann Arbor.

In August 2006, he became an assistant professor in the Department of Statistics at Carnegie Mellon University in Pittsburgh

[japanese anime]

Sounds like a physicist, not a psychologist.

Plus I don't expect someone with a last name like "Shalizi" to know what he's talking about.

[Soulless Doubter]

Sounds like a physicist, not a psychologist.

Plus I don't expect someone with a last name like "Shalizi" to know what he's talking about.
I have yet to read the article and am admittedly prejudiced against his side of the argument at this point, but the man is an instructor of statistics at CMU. That alone means he has the technical ability to fully understand the topics in psychometrics.

Your second sentence is meaningless, even if you were joking.

[Angler]

Do you expect someone who calls it "principle component analysis" instead of "principal component analysis" to be well versed in psychometric literature?The author calls it "principal component analysis" in the article as well, so the use of "principle" is just a minor spelling error akin to a typo. Even excellent English speakers can make simple mistakes like that from time to time.

Sounds like a physicist, not a psychologist.Obviously the author has a strong background in statistics and is very well qualified to understand the factor analysis on which the concept of g is based.

[Ahknaton]

Sounds like a physicist, not a psychologist.

Plus I don't expect someone with a last name like "Shalizi" to know what he's talking about.
Do you have any criticisms of the article itself, besides pointing out a spelling mistake? You don't need to be a psychologist to critique the statistical arguments for g.

Anyway, I thought he made a good argument. Note that he is not criticising the validity of psychometric tests per se (since a battery of separate tests can measure specific mental abilities), he is simply arguing against the idea of a single general intelligence factor.

[Saqqara]

Trying to avoid the obvious intuitive thought that people who are against IQ tests and classifications didn't do as well as they had hoped they would, I agree that smart people often do stupid things. I'll venture to ask the following questions.

What is the nature of intelligence?

What causes intelligence or the lack thereof?