On Nov 17, 2005, at 2:08 PM, Jason Palmer wrote:

> It is my understanding that Bayesians also believe in fixed  
> distributions for given phenomena.

At least one major proponent of Bayesian methods -- E. T. Jaynes --  
would strongly disagree with this statement, calling it an example of  
the "mind projection fallacy" [1].  Jaynes views probability theory  
as an extension of classical logical that can handle degrees of  
certainty between the two extremes "certainly true" and "certainly  
false," a view that is supported by Cox's Theorem [2,3].  Thus a  
Bayesian
probability distribution represents only one's state of knowledge  
about some aspect of the physical world -- it is not a property of  
the physical world itself.  The notion of "imprecise" probabilities  
is then nonsensical, as presupposes that there is some "correct", but  
unknown, distribution that is a property of some physical system.

1. E. T. Jaynes, Probability Theory: The Logic of Science.  Cambridge  
University Press, 2003.

2. R. T. Cox, "Probability, frequency, and reasonable expectation."  
American Journal of Physics 17 (1946), 1--7.

3. K. S. Van Horn, "Constructing a logic of plausible inference: a  
guide to Cox's Theorem." International Journal of Approximate  
Reasoning 34, no. 1 (Sept. 2003); available on the web at http:// 
www.leuther-analytics.com/bayes/papers.html.

_______________________________________________
uai mailing list
[email protected]
https://secure.engr.oregonstate.edu/mailman/listinfo/uai

Reply via email to