correlation is not causation

in response to this post at GNXP, I left a comment expressing my skepticism that the mean IQ variation by country is meaningful. GC answered:

how the methodology for basing any argument on IQ manages to extricate the confounding effects of economic class.

Hmmm, Aziz - your argument seems to be that your parents' economic status determines how well you do on IQ tests. But we've blogged before on the result that (for example) your IQ correlates much more highly with *your* economic status than with your parents' economic status.

Valid point, but ignoring the fact that *my* economic status correlates with my parents'. GC mentions some anecdotal evidence in order to claim that immigrants from poorer countries do better on IQ tests, but I think that the barrier to immigration is itself an outbound filter - it takes a special breed of person to immigrate to the US in search of a better life, and those kinds of people tend to work a lot harder in pursuit of that goal than, say, indigenously-raised spoiled suburbanites. Often the children of those selfsame immigrants, mind you.

On the whole, there's far too much reliance on correlation as a proxy for causation. For example, GC links to some recent studies that suggest a correlation between IQ and brain volume - which is such a tenous foundation for the kinds of social analysis that GNXP performs much more robustly with genetic data, that I'm going to have to get off my rear and blog further about this gross misuse of MRI (especially fMRI) over at Reference Scan. But for a preview of the argument I shall make, consider the correlation between physical strength and the density of fast-twitch muscle fiber.

No comments: