Abstract

The paper investigates what are the proper procedures for calculating the probability on certain evidence of a particular object e having a property, Q, e.g. of Eclipse winning the Derby. Let `α ' denote the conjunction of properties known to be possessed by e, and P(Q)/α the probability of an object which is α being Q. One view is that the probability of e being Q is given by the best confirmed value of P(Q)/α . This view is shown not to be generally true, but to provide a useful approximation in many cases. Then given that we have information about the observed frequencies of Q among objects having one or more of the properties whose conjunction forms α , the paper shows how to establish which value of P(Q)/α is the best confirmed one