On 27 Nov 1999 18:44:33 -0800, [EMAIL PROTECTED] wrote:
> Obviously the sets are not related in a linear fashion.
>
> I would suggest that a 4th degree polynomial equation best fits the data.
Oh! that should have been obvious....
> The average percent error of the residuals is 2.3948% with a standard
> deviation of 1.8634%, the maximum percent error is 5.7357%. A sixth degree
< ... >
( I guess I missed the part where the scores had a natural zero --
unlike IQ or SAT scores -- or this would be silly advice.)
>
> A good statistical software package would advise you that a linear plot is no
> good and suggest trying curvilinear (polynomial) fits. It took me about five
> minutes to check for linear correlation and also check 2nd, 3rd, 4th, 5th,
> and 6th degree polynomials.
- and do keep in mind, that free advice from a stat-package is worth
about what you paid for it. (Especially, *that* piece.)
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
