I didn’t mean to rehabilitate frequentism! I only meant to point out that calibration is a frequentist optimality criterion, and that it’s one that Bayesian posterior intervals can be proved not to have in general.
Too late. I have already updated to believe that a theory that demands priors can’t be complete. Correct, maybe, but not complete. We should work out an approach that works well on more criteria instead of guarding the truth of what we already know.
If Bayes were the complete answer, Jaynes wouldn’t have felt the need to invent maxent or generalize the indifference principle. That may be the correct direction of inquiry.
ETA: this was a response to Cyan saying he didn’t mean to rehabilitate frequentism. :-)
I didn’t mean to rehabilitate frequentism! I only meant to point out that calibration is a frequentist optimality criterion, and that it’s one that Bayesian posterior intervals can be proved not to have in general.
Too late. I have already updated to believe that a theory that demands priors can’t be complete. Correct, maybe, but not complete. We should work out an approach that works well on more criteria instead of guarding the truth of what we already know.
If Bayes were the complete answer, Jaynes wouldn’t have felt the need to invent maxent or generalize the indifference principle. That may be the correct direction of inquiry.
ETA: this was a response to Cyan saying he didn’t mean to rehabilitate frequentism. :-)
Updated, eh? Where did your prior come from? :)
Overcoming Bias. :-)