Re: Induction of statistical models

From: George Kahrimanis (anakreon_at_hol.gr)
Date: 12/24/04 

Date: 23 Dec 2004 16:02:33 -0800

Herman Rubin wrote:
>There is no basic disagreement between Bayesian inference
>and Occam's razor, if the Bayesian inference is done
>properly, not rashly.

Being a nonexpert, I just suspect that the point of view in the
following article is in that ballpark.

"Statistical Inference, Occam's Razor and Statistical Mechanics
on The Space of Probability Distributions"
by Vijay Balasubramanian, in <http://www.arXiv.org>
(either in the "physics" or in the "math" archive)
and in other websites.

>>3. On the Epicurean principle "It would be unscientific to choose an
>>arbitrary hypothesis if several are consistent with the data", and a
>>possible synthesis with Occam's razor:
>>http://www.stat.columbia.edu/~cook/movabletype/archives/2004/12/wacky_computer_1.html

I seems to me that I am an Epicurean wrt this principle. Yet I
detest infinities; therefore I am happiest with a small bunch of
models, to minimize forseeable trouble (all things considered).

(In a similar vein Aristotle instructs that maximum happiness
requires that you keep a very small number of friends.)

Many thanks for the references. I am sure that I shall enjoy
reading them, whether on model building or model selection.
~ George Kahrimanis