Next: The Likelihood Maximizes
Up: 3: Estimation of
Previous: Interpretation of Bayes
- Natural parameterization of ignorance is
constant.
- But suppose we put
and then 
has a
priori probability 
- So the concept of ``complete ignorance'' is not well defined, as
change of variables changes concept.
Jeffreys (a big bad English-Oxford University Press
book cited in Mathews and Walker) spends 100's of pages trying to avoid
this difficulty.
His efforts are, in my opinion, so much garbage.
- Fortunately, there is no trouble for large n because of both an
obvious numerical reason (see next slide (The Likelihood Maximizes) and
a not so obvious mathematical reason (see slide (General Maximum
Likelihood Methods)).
Geoffrey Fox, Northeast Parallel Architectures Center at Syracuse University, gcf@npac.syr.edu