Re: Race, intelligence, and anti-racist prejudice (Was: Genetic Evolution)
Stephen Lajoie (email@example.com)
Sat, 4 Feb 1995 07:11:15 GMT
In article <2FEB199515064125@rosie.uh.edu>,
JAMES BENTHALL <firstname.lastname@example.org> wrote:
>For one thing, there has never been a test that I'm aware of that involved
>"billions of samples." If the variation between groups is much less than
>the variation within groups what enables us to say that they are *two
>Part of the confusion is my misstatement in an earlier post. (I never claimed
>to be above bullshit! Just ask readers of the "What is Natural" thread :-)
>I said t-test but what the professor actually said was the "analysis of
>variance" tests (ANOVA), but t-tests also analyze variance. Differences
>_between_ means are usually swamped by the variation _within_ populations.
>(Hopefully,everyone can understand this statement.)
ANOVA test are usually applied to small sample sizes, where the variation
within a populations may very well "swamp" the variation between
populations. It is more likely you would reach the wrong conclusion if
you have a small sample size. For example, if the sample size was 3 in
each population, there is far less certainty because you may get someone
from one extreme or the other in your sample that is very atypical of a
population of 3.
One way to resolve this is to increase sample size.
Note that the sample size in these IQ test between races was in the
hundreds or thousands, making it very easy to resolve the different
curves with great certainty.
>Standardized intelligent qoutient tests are usually felled by the analysis
>of variance. Differences between populations are so much greater than
>differences within populations that it makes it rather difficult to say
>generalizations such as "Population A tests consistently above that of
>population B." As a result, folks engaged in these kind of tests don't
>run ANOVA tests.
ANOVA test would not be a meaningful test with such large samples. For
one thing, the variance is not the issue so much as the means. Secondly,
the large sample sizes used in most of the studies would produce values
of F that pretty much reduce to _What you see is what you get_. That is,
you are very sure of your results.
>I have no references, only the reassurance of an esteemed professor of
>anthropology that taught my stats class. If you want the truth you must
>INVESTIGATE. I, frankly, don't have time.
Yes, but, being bored as I am, I did! I dug out my stat book from
psychology (kinda worthless) and my stat book from engineering (ah! I
Steve La Joie