Re: Bayesian Networks and Belief Functions

[Kevin S. Van Horn]

Rolf Haenni wrote:
> (even to speak about "probability that God exist" is absurd).

Only for staunch frequentists.  The dogmatic insistence of frequentists that
probabilities are limiting frequencies, and *only* limiting frequencies, has led
them to a greatly impoverished probability theory whose range of applicability
is unnecessarily limited.  This has lead to numerous convoluted devices to solve
problems that are easily and naturally solved once one is willing to risk
eternal damnation by taking the heretical step of assigning probabilities to
non-repeatable events.

This is why I'm such an advocate of Bayesian methods.  Quite apart from Cox's
Theorem, there is the purely practical consideration that Bayesian methods give
straightforward answers to problems that people tie themselves into knots trying
to solve with other approaches.  Just watch a classical statistician (whose
methods are based on frequentist doctrine) run screaming from you when you
mutter, "nuisance variable."