Alcune consideraziooni sulla quantità d'informazione relativa a un esperimento

Abstract

An experimental result may affect pre-experimental belief in two ways: it can agree or disagree with experimental belief. Such aspects were considered using two concepts of information provided by an experiment: the first due to Lindley, the second to Kullback-Leibler. It can be seen that the Lindley's information is in favour of results which agree with prior belief, while the Kullback-Leibler's information considers more useful results which disagree with prior belief. Finally it is noticed that the maximum likelihood estimator maximizes the Kullback-Leibler information.