From: Iain Strachan (iain.strachan@eudoramail.com)
Date: Wed Nov 22 2000 - 20:31:13 EST
On Thu, 21 Nov 2002 13:42:36 Dr. Blake Nelson wrote:
>Glenn, et al.
>
>I am going to take a wild stab here on something that
>I have read absolutely nothing about other than part
>of what has gone back and forth between you and others
>on the list. I know that you assert that Dembski
>cannot detect design, because any random sequence can
>be made to be meaningful by coming up with a post hoc
>code to make it say something meaningful. This may or
>may not be detrimental to Dembski's project.
>
>In looking for relationships among data I can ALWAYS
>come up with an equation that matches all the data
>points EXACTLY. This is nothing unusual. The problem
>is, in finding an equation that actually describes the
>relationship (if one exists). If I am writing the
>equation to specifically hit all the data points
>(rather than finding a general relationship among the
>data), I use up all my degrees of freedom and the
>results are not going to be statistically significant.
> Thus, my solution is obviously ad hoc.
>
>My strong intuition is that there is an easy way to
>distinguish between something post hoc and clumsy like
>your "code" and something that appears to be designed.
> If your code were an equation, I would easily reject
>it as not likely to be design, because I have no
>degrees of freedome left, so to speak.
>
>
>
This seems to be absolutely spot on. The same argument applies in my
field of neural networks. If you give me a set of 10 points, and I
fit a 10th Degree polynomial through them, by least-squares fitting,
I will ALWAYS get an exact fit. This means that I cannot determine
whether the data was designed or just random. You may come along and
tell me that you generated the data points from a 34th degree
polynomial and that I've failed to detect your design, but that's
just because there aren't enough points to detect design.
Iain.
Join 18 million Eudora users by signing up for a free Eudora Web-Mail
account at http://www.eudoramail.com
This archive was generated by hypermail 2.1.4 : Sat Nov 23 2002 - 00:39:31 EST