Objective Bayesian Statistics

Abstract

Bayesian inference --- although becoming popular in physics and chemistry --- is hampered up to now by the vagueness of its notion of prior probability. Some of its supporters argue that this vagueness is the unavoidable consequence of the subjectivity of judgements --- even scientific ones. We argue that priors can be defined uniquely if the statistical model at hand possesses a symmetry and if the ensuing confidence intervals are subjected to a frequentist criterion. Moreover, it is shown via an example taken from recent experimental nuclear physics, that this procedure can be extended to models with broken symmetry.