[cljs-js-eval]

One of his repeated criticisms of analysis like this is using regression to fit datasets that don't necessarily follow a straight trend line.

His larger point with respect to IQ is that it only predicts substandard IQs. Measures like the variance are largely influenced by the effects of IQs below the average, not above the average.

A second point which I don't see addressed is that IQ measures share features with testing. If how you test in school affects future income, and testing well in school predicts testing well on IQ, then more robust analysis is needed. A sample space of people in a wide band of IQ ranges who all got about the same grades in school would be a true test of the predictive value of IQ on income.

[reasonaway]

it is all unfounded speculation based on mathematica simulations and his own assertions. there is no empirical support for his arguments.