Several known procedures transforming an imprecise probability into a precise one focus on special classes of imprecise probabilities, like belief functions and 2--monotone capacities, while not addressing the more general case of coherent imprecise probabilities, as defined by Walley. In this paper we first analyze some of these transformations, exploring the possibility of applying them to more general families of uncertainty measures and evidencing their limitations. In particular, the pignistic probability transformation is investigated from this perspective. We then propose a transformation that can be applied to coherent imprecise probabilities, discussing its properties and the way it can be used in the case of partial assessments.
Publisher's on-line resources