Random sets as imprecise random variables (Q555822)

The authors study relationships between the family of probability measures dominated by the upper probability associated with a random set and the family of distributions of measurable selections of this random set. In particular, they prove that if the upper probability is attained by the distributions of selections (i.e. the upper probability is coherent), then the weak closures of the family of dominated probabilities and the family of distributions of selections coincide. The new features of this result are the lack of completeness assumption on the carrier (separable metric) space and the fact that random sets are not necessarily closed.