I wanted to thank and congratulate Stefan Molyneaux for his dogged investigation of the career killing topic: Race, Sex and IQ.
I think he has finally proposed a hypothesis that is worth researching and is provable and scientific.
The hypothesis I see as valid is:
(1) There are differences in mean IQ by race and gender.
(2) owing to the shape of the normal distribution, the proportion of people occupying a given interval of IQ scores will vary with distance from mean. (example: the interval of IQ 100-115 contains 34% of people but the interval of 130-145 contains 2% of people)
(3) TO BE PROVEN: When dealing with the middle 68% of individuals (+/- 1 std, the people you most often interact with) there is less confidence that racial/sex differences in IQ will be observed owing to variations in and between groups. (This can be shown with a power calculation)
(4) MOLYNEAUX'S INSIGHT: For an equal size IQ interval on the tails of a normal distribution, the confidence about IQ increases nonlinearly. SO given some prior knowledge, (e.g. that a person has an IQ>115, i.e. a so called 'elite') once can be more confident of observing bias in race/sex proportions.
At the very least statistical tests can be done to prove that one needs a larger sample of people from IQ[85,115] interval to be confident of a race/sex difference than one would need from an IQ[130,160]
It needs some clean up but it seems like a good paper if it hasnt already been published.
If the work has been done I would be interested in a citation.