I mean Bayesian reductionist evidence. Which means, anecdotes count, even if they still count for less than numbers and double-blind tests.
I think this is a misunderstanding of the correct application of Bayes’ Theorem.
That a comment opening with this quote-reply pair is voted above zero troubles me. It is a direct contradiction of one of the most basic premises of this site.
I would have voted it down were it not for the rest of the paragraph cited, which basically comes down to “anecdotes are Bayesian evidence, but with caveats related to the base rate, and not always positive evidence”. Which is, as best I can tell, correct. In isolation, the opening sentence does seem to incorrectly imply that anecdotes don’t count at all, and so I’d have phrased it differently if I was trying to make the same point, but a false start isn’t enough for a downvote if the full post is well-argued and not obviously wrong.
In context, I interpreted pjeby to be saying that anecdotes counted as evidence which should lead a Bayesian rationalist to believe the truth of PUA claims. If that was not their intention I got them totally wrong.
However if I interpreted them correctly they were indeed applying Bayes incorrectly, since we should expect a base rate of PUA-affirming anecdotes even if PUA techniques are placebos, and even in the total absence of any real effects whatsoever. It’s not evidence until the rate of observation exceeds the base rate of false claims we should expect to hear in the absence of a non-placebo effect, and if you don’t know what the base rate is you don’t have enough information to carry out a Bayesian update. You can’t update without P(B).
That a comment opening with this quote-reply pair is voted above zero troubles me. It is a direct contradiction of one of the most basic premises of this site.
I would have voted it down were it not for the rest of the paragraph cited, which basically comes down to “anecdotes are Bayesian evidence, but with caveats related to the base rate, and not always positive evidence”. Which is, as best I can tell, correct. In isolation, the opening sentence does seem to incorrectly imply that anecdotes don’t count at all, and so I’d have phrased it differently if I was trying to make the same point, but a false start isn’t enough for a downvote if the full post is well-argued and not obviously wrong.
In context, I interpreted pjeby to be saying that anecdotes counted as evidence which should lead a Bayesian rationalist to believe the truth of PUA claims. If that was not their intention I got them totally wrong.
However if I interpreted them correctly they were indeed applying Bayes incorrectly, since we should expect a base rate of PUA-affirming anecdotes even if PUA techniques are placebos, and even in the total absence of any real effects whatsoever. It’s not evidence until the rate of observation exceeds the base rate of false claims we should expect to hear in the absence of a non-placebo effect, and if you don’t know what the base rate is you don’t have enough information to carry out a Bayesian update. You can’t update without P(B).