Representation insensitivity in immediate prediction under exchangeability (Q962847)

The authors consider the problem of representation insensitivity in immediate prediction, where a subject, using a number of observations of a finite number of exchangeable random variables, is asked to coherently model his beliefs about the next observation, in terms of a predictive lower prevision. They study when such predictive lower previsions are representation insensitive, meaning that they are essentially independent of the choice of the (finite) set of possible values for the random variables. The authors define predictive systems, and the properties they impose (exchangeability and representation insensitivity) lead them to consider mixing predictive systems and, more specifically, IDMM-predictive systems. They establish that such representation insensitive predictive models have very interesting properties, and show that among such models, the ones produced by the imprecise Dirichlet-multinomial model are quite special in a number of ways. In the conclusion, they discuss the open question as to how unique the predictive lower previsions of the imprecise Dirichlet-multinomial model are being representation insensitive. As they explain, much more work is needed, however, to be able to draw a complete picture of the issue of representation insensitivity in predictive systems.