Bayesian Inference

Bayesian inference treats an unknown quantity as something you hold beliefs about, and uses data to update those beliefs. You start with a prior (先驗) — what you believed before seeing data — combine it with the likelihood of the data, and get a posterior (後驗): your revised belief afterward.

Its appeal is that it answers the question people actually want — “given my data, how probable are different values of the parameter?” — which a p-value pointedly does not. The trade-off is that conclusions depend on the prior you chose, so a transparent, defensible prior matters; with lots of data the prior fades and Bayesian and classical answers tend to converge.