Kishkumen wrote: ↑Fri Jun 17, 2022 2:03 am
Bayesian analysis does not make good historical arguments better or bad historical arguments better, as far as I can tell. Maybe it makes arguments look better because people find equations and numbers so impressive these days.
I can't say that I've seen many applications of Bayesian inference in history, but I think this simply has to be true, given what Bayesian inference is. It is literally nothing—nothing at all—but simple logic. Bayes's theorem consists of two lines: an obvious statement of logic followed by a division.
If you have solidly quantitative data then Bayesian inference lets you express its logical implications in the quantitative terms that solidly quantitative data can actually support. It avoids wasting a lot of valuable quantitative information by crudely translating it into vague, qualitative terms before drawing conclusions from it. This is obviously great.
When all your probabilities are simply made up subjectively, though, then Bayesian inference cannot possibly provide anything that careful reasoning in ordinary language would not also provide—except perhaps a false confidence. And I'm afraid that crude efforts at counting sparse, incomplete, and incommensurable data points do not actually amount to anything more than making up probabilities subjectively.
Bayesian inference is an important methodology
in statistics. History has sample size one.
I was a teenager before it was cool.