3 ms·

> IQ tests don't even have validity well outside 2 s.d. An easy way to measure above +2 z-scores is to administer an IQ test at a much earlier age than usual.

> IQ tests don't even have validity well outside 2 s.d. An easy way to measure above +2 z-scores is to administer an IQ test at a much earlier age than usual. That's what the famous Study of Mathematically Precocious Youth did. It has proven to be super effective even at the extreme right tail of the curve. [1] http://4.bp.blogspot.com/-4b8ThyfDgzs/U29nQEBAQTI/AAAAAAAAE40/iu0AgfuHE8E/s1600/smpy_odds.png http://4.bp.blogspot.com/-4b8ThyfDgzs/U29nQEBAQTI/AAAAAAAAE4... > I believe you're incorrect regarding the variance in male IQ vs female. Belief doesn't enter into it. Many studies report higher male variance and it is widely known in the psychological community. While people try and attack this as a recruitment pattern problem (see [2]) or try to prove that it is partly cultural (see [3]), only people unfamiliar with the literature would outright deny it. [2] http://www.sciencedirect.com/science/article/pii/S0160289608000962 http://www.sciencedirect.com/science/article/pii/S0160289608... [3] http://www.sciencedirect.com/science/article/pii/S0160289610000346 http://www.sciencedirect.com/science/article/pii/S0160289610...

nl 10y ago
An easy way to measure above +2 z-scores is to administer an IQ test at a much earlier age than usual. Your graph shows that this study administered a SAT Math test at a much earlier age, not an IQ test. The study itself says: "Before age 13, these participants scored within the top 0.01% for their age on either SAT mathematical reasoning ability (SAT-M 700) or SAT verbal reasoning ability." I'm unconvinced that this shows either way if "IQ tests [don't] even have validity well outside 2 s.d" (I agree with the evidence of gender variance in IQ though. An unpaywalled copy of the second article is at https://tip.duke.edu/about/research/intelligence_article.pdf https://tip.duke.edu/about/research/intelligence_article.pdf an is interesting reading.)
DaniFong 10y ago
To be clear, I think that the stated belief that there is higher male variance is possibly true. I think it has a good chance of being false, but that's not the point. Statistically, for example, the male population has a greater chance of genetic disease and indeed lives shorter life; this is supposed to be in part because of the lack of two x-chromosomes, which, as a mosaic, both express themselves in the ordinary female body. What's emphatically not true is assuming a normal distribution -- specifically out to the tails. It's not even clear that the variance doesn't turn around at some point to have higher female variance at the top end of the curve. It is extremely easy to synthesize violations of a normal distribution of ability for female subjects -- there seems to be a powerful threshold effect that, in addition to cutting out large numbers of young women in certain cultures in mathematical competition, tends to preferentially spare the very top. If things were normally distributed, of those women that compete in mathematics competitions, almost none would make it to the very top of the competition, mathematically. That has been proven empirically false. Few scholars directly assert that the normal distribution is a perfect model for intelligence, but people seem to take the implications quite seriously. I'm pretty sure they're mistaken on that front.

nl 10y ago

An easy way to measure above +2 z-scores is to administer an IQ test at a much earlier age than usual. Your graph shows that this study administered a SAT Math test at a much earlier age, not an IQ test. The study itself says: "Before age 13, these participants scored within the top 0.01% for their age on either SAT mathematical reasoning ability (SAT-M 700) or SAT verbal reasoning ability." I'm unconvinced that this shows either way if "IQ tests [don't] even have validity well outside 2 s.d" (I agree with the evidence of gender variance in IQ though. An unpaywalled copy of the second article is at https://tip.duke.edu/about/research/intelligence_article.pdf https://tip.duke.edu/about/research/intelligence_article.pdf an is interesting reading.)

DaniFong 10y ago

To be clear, I think that the stated belief that there is higher male variance is possibly true. I think it has a good chance of being false, but that's not the point. Statistically, for example, the male population has a greater chance of genetic disease and indeed lives shorter life; this is supposed to be in part because of the lack of two x-chromosomes, which, as a mosaic, both express themselves in the ordinary female body. What's emphatically not true is assuming a normal distribution -- specifically out to the tails. It's not even clear that the variance doesn't turn around at some point to have higher female variance at the top end of the curve. It is extremely easy to synthesize violations of a normal distribution of ability for female subjects -- there seems to be a powerful threshold effect that, in addition to cutting out large numbers of young women in certain cultures in mathematical competition, tends to preferentially spare the very top. If things were normally distributed, of those women that compete in mathematics competitions, almost none would make it to the very top of the competition, mathematically. That has been proven empirically false. Few scholars directly assert that the normal distribution is a perfect model for intelligence, but people seem to take the implications quite seriously. I'm pretty sure they're mistaken on that front.