Commenter Neandercel said the following about the -0.08 correlation between IQ and weight/height ratio:
PP, please come to the southern US for a month.
-0.08 seems unbelievably low to me. I wonder whether the correlation would have been stronger had waist:height ratio (or some combined index of waist:height ratio and waist:hip ratio) been used as a proxy for corpulence in place of BMI.
Given that I derived my -0.08 figure from a study with over 6000 people, I’m pretty confident in it and that confident is affirmed by data from Kanazawa (2013) (bottom of page 438) who also used massive samples to show the correlation between IQ and BMI in adulthood ranged from -0.096 to -0.124.
In other words, a correlation of about -0.1 is well replicated, at least in the UK.
However there is reason to think these correlations are underestimates. The -0.08 figure I calculated came from a study which used the NART as a proxy for IQ however the NART only correlates 0.69 with the gold standard WAIS-IV. To crudely correct for this, let’s divide -0.08 by 0.69, strengthens the correlation to -0.12.
The other problem is BMI data is self-reported and people lie about their height and weight. In one study, the Pearson correlation between self-reported and objectively measured BMI was 0.75. Correcting for this further strengthens the correlation to -0.16.