From: Iain Strachan (iain.strachan@eudoramail.com)
Date: Fri Dec 15 2000 - 19:59:29 EST
Glenn,
>>Generate 100 pairs of (x,y) points from random numbers in the
>>range 0-1 (this can be done with the Excel RAND() function. Add a
>>101st (x,y) pair and make it equal to (100,100). Now compute the
>>correlation coefficient between the two sequences (using the Excel
>>CORREL() function). You will get an answer for R that is close to
>>0.999. So your "objective math" is telling you that the sequences
>>are highly correlated.
>
>So???? They are highly correlated.
So you're telling me that a unit square full of random numbers plus
the point at (100,100) looks correlated? No a single statistician
worth his/her salt would support that view. It would be regarded as
an abuse of statistics.
I see nothing here to support your contention
>that R is not a good measure of correlation at all.
>
Didn't you read the quote from Press et. al? This isn't _my_
contention; it's straight from a standard text book that you will
find in virtually any university computer science department in the
world. The assertion is that R is only a good measure of the
strength of correlation if you already know that a correlation exists
& that boils down to a knowledge of what the probability
distributions of the variables look like.
The remainder of your argument about sequences of bases not
containing outliers is irrelevant. I never said that they could or
could not contain outliers. The outlier demonstration was simply a
specific example I gave in response to your point about correlation
coefficients, but as usual you try to shift the ground of the
argument.
We could go down the route of talking about HMM's to analyse DNA
sequences & relate that to Minimum Description Length etc, but I'm
not interested in engaging in a long point by point debate about
whether Dembski can detect design or not; you have evidently decided
a priori that he cannot; I on the other hand remain open minded and
continue to look for ways in which his ideas relate to my
understanding of machine learning, which has been my academic area of
research for the last seven years. Rest assured, if I find something
in Dembski that can easily be refuted and shown to be illogical,
within the terms of statistical inference, I'll waste no time in
shouting about it. I've not found anything yet, and nothing you have
said has changed my mind; on the contrary, it has focused my
understanding, that there is nothing wrong with the basic maths. It
may be that it needs to be developed somewhat; or perhaps shown in
relation to existing techniques (I have identified right fr!
om
the start that MDL is the clear way forward), but I don't find
anything fundamentally wrong at the moment.
Iain.
Need a new email address that people can remember
Check out the new EudoraMail at
http://www.eudoramail.com
This archive was generated by hypermail 2.1.4 : Sun Dec 15 2002 - 23:42:34 EST