On 1/15/08, dlakelan <dlakelan at street-artists.org> wrote:
> As far as I can see, Bayesian methods are the future of statistics. if
> maxima can become strong in this one area due to its useful symbolic,
> and numerical combination, that would be great.
Bayesian inference is an excellent example of a problem domain
which can benefit from mixed symbolic and numerical methods.
The game, as I see it, is to express or represent the problem in
Maxima, apply some well-known rules, return a symbolic result
if that's what the rules yield, otherwise carry out a numerical
approximation. Bayesian inference is much more regular than
frequentist stuff, so it is especially susceptible to automation.
Many practical problems can be expressed as belief (probability)
or value (utility) calculations. If Maxima were equipped to handle
probability and utility calculations, a wide range of practical
applications would result.
FWIW
Robert