L'INFERENZA STATISTICA SECONDO BAYES, FISHER E NEYMAN-PEARSON

Abstract

In the literature about procedures of statistical inference, which – on the basis of a random sample – allow to accept a precise hypothesis concerning the probability distribution which yields the sample, there are many proposals. One of the oldest is the one by T. Bayes (1763): starting from a given distribution for the parameter of a random variable, this same distribution is revised conditionally on the sample results. Instead, R. Fisher’s proposal is based only on the sample information: a given hypothesis is rejected only if the sample results are scarcely coherent with such hypothesis. The third proposal, made by J. Neyman and E. Pearson, takes into explicit consideration also an alternative hypothesis, and leads to the acceptance of the hypothesis more akin to the sample results.