Bayes’ postulate for trinomial trials
Márcio Alves Diniz, Adriano Polpo · AIP conference proceedings · 2012
In this paper, we discuss Bayes’ postulate and its interpretation. We extend the binomial trial method proposed by de Finetti [1] to trinomial trials, for which we argue that the consideration of equiprobability a priori for the possible outcomes of the trinomial trials implies that the parameter vector has Dirichlet(1,1) as prior. Based on this result, we agree with Stigler [2] in that the notion in Bayes’ postulate stating “absolutely know nothing” is related to the possible outcomes of an experiment and not to “non-information” about the parameter.