B1700
Title: The risk of random sets with applications to basket derivatives
Authors: Christian Gourieroux - University of Toronto and CREST (Canada) [presenting]
Abstract: The risks are analyzed in random sets and their implications for basket derivatives. Based on an extension of integration by parts for random sets, stochastic dominance of orders 1 and 2 for random sets is defined. Since the ordering of sets, that is the inclusion, is a partial order, left and right notions of stochastic dominance are distinguished. The observed sets are in a one-to-one relationship with observed multivariate binary variables, each component of which indicates high or low risk for a given type of risk. This relationship is used to define basket derivatives and to develop statistical inference. The special cases of exchangeability, the law of determinantal point process (LDPP), local pairwise interactions and block models are considered for illustration.