Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!rutgers!umnd-cs!umn-cs!umnstat!weiss
From: weiss@umnstat.UUCP (Robert Weiss)
Newsgroups: sci.bio
Subject: IQ, if any (was:Re: Knowledge and the Academics)
Message-ID: <337@umnstat.UUCP>
Date: Sun, 21-Jun-87 19:46:47 EDT
Article-I.D.: umnstat.337
Posted: Sun Jun 21 19:46:47 1987
Date-Received: Tue, 23-Jun-87 00:47:11 EDT
Organization: Statistics, U. of Minn, Mpls, Mn
Lines: 23
Keywords: IQ, Gaussian, Laplacian, Normal distribution.


-  From: eddy@boulder.Colorado.EDU (Sean Eddy)

-But IQ scores are distributed on that classical (artifactual?)
-Gaussian curve. 

Artifactual.  IQ _TEST_ scores might possibly be distributed APPROXIMATELY 
as a bell-like curve, but only by choice of the test designer, NOT by 
anything to do with IQ.  

Let us suppose that in a particular population, and with a particular 
test, that the test scores are well-defined.  Let F(x) be the percent 
of people in the population scoring less than x on the
test.  Let us suppose that F(x) is approximately Gaussian.  To
get any other shape that you wish, take new "GIQ" scores to be equal to 
  GIQ(x) = G(F(x))  
With careful selection of the function G, you can get GIQ scores to 
approximate any shape that you like: semicircle, chisquare, or
uniform (G=identity).

Robert Weiss
ihnp4!umn-cs!umnstat!weiss
umnstat!weiss@umn-cs.ARPA