I'm familiar with IQ, and, though to a more non-technical degree, bell curves.
A. "Select any parts ..."
"Marijuana users" and "non-marijuana users" are not "any"; furthermore, this is only if you calibrate the test answers->IQ to each group individually. This would be a patently stupid idea, and is not the one implied by the use of an IQ test.
By your arguments, if we were to take a random sample of the mentally retarded ( using the IQ standard, i.e. technically our sample is entirely composed of those with an IQ below 70 ) , and compared it to a random sample of people in general, we would get the same distributions.
B. "Note that I have said random." You didn't state random
samples; you stated that "a random distribution... forms a bell curve."
If you use a selection bias with random selection, would would get different, but similar bell curves.
You stated "similar" for your own case, also.
And the only way to know if this degree of "similar"ity is within statistical insignificance, one would have to actually perform the tests.
I don't even know why I'm saying as much as I am. It all comes down to the fact that we are talking about two different groups of people who could, possibly, exhibit different intelligence distributions for a variety of reasons, the only way to necessarily know will not be the case before seeing results being to intentionally and very precisely control the samples ( i.e. non-random ) , or to, as I stated before, calibrate the IQ results for each group individually, instead of using the same scale for the entire group of test-takers- which, as I also stated before, is absurd and rediculous for such purposes as testing how marijuana affects intelligence or how users of it tend towards some abnormal degree of intelligence.
None.