"Mark D. Morin" <mdmpsyd at NOSPAMgwi.net> wrote in message
news:3D2D4B4C.9040009 at NOSPAMgwi.net...
> John Knight wrote:
> > Where? You keep saying this, and I keep pointing out that any possible
> > measurement error is 1/20th to 1/70th of the sex differences by race,
and
> > you come back again with the same comment about measurement errors.
What
> > exactly are you referring to?
>> Try taking a look at HOW they computed these statistics.
>
ok, so you are just going to keep denying it in the hopes that people will
accept what you say? Measuring brain size to a high degree of accuracy is
well within the capabilities of this technological society, so your denial
of the accuracy, coupled with your failure to state specifically which data
you believe is suspect, means nothing.
> > uh, how much, where's the url, how does it compare to the sex and race
> > differences, could this explain why correlation is .88 rather than 1.0?
>> No URL. Real Journals, Real research. Pull any issue of Neurology.
> There is a VERY robust correlation to brain volume and age. Any
> research that does not attempt to control for this fact is illustration
> of the fact that the individuals do not know how to conduct research.
>
This is not a true statement. Even IF there is "a VERY robust correlation
to brain volume and age", it's simply not true that this MUST be a
consideration in the final correlation.
It's particularly not true if correlation between brain size and GRE scores
is already as high as 0.8795 even without correcting for age.
In fact, if the data were to be adjusted for age before the final average
brain size by race and sex were calculated, that could [and definitely
would] be used as an argument for why the data *may* have been manipulated
or biased.
The safest, most conservative approach is to use the data which is *not*
adjusted for age in the correlation, and then compare that to a correlation
which uses the data *after* it's adjusted for age. The likely scenario
would be that r-squared would *increase* to more than 0.8795, which isn't
even necessary to demonstrate Peter's point.
Adjusting it for age first raises too many questions about this already
allegedly "controversial" issue.
However, since you didn't even take a stab at explaining how failing to
adjust for age would impact the correlation, your argument is specious from
the beginning. The subjects were aged 20 to 90 and STILL correlation was
high. Perhaps it's true that it would have been higher if all subjects had
been 39 years old, but so what? How does that change the significance of
the correlation? It doesn't.
Let's use an outlandish assumption. Let's assume that the brain size of
Black women increases by a standard deviation between 20 and 89 years of age
(an obviously ludicrous assumption), from the median of 1,217 cc's to 1,336
cc's. This would still be a standard deviation lower than the median for
Asian men, and two standard deviations lower than the upper end of Asian
men.
IF Black women are being under-reported because they had such a massive but
unknown increase in brain size over 69 years, and their actual average brain
size over that time was 1,267 rather than 1,217 cc's, not even this would be
enough to change r-squared by that much.
So your concern that adjusting for age is a necessary first step is dead
wrong.
John Knight
