A1386
Title: Classes of high dimensional Bernoulli distributions and applications
Authors: Patrizia Semeraro - Politecnico di Torino (Italy) [presenting]
Roberto Fontana - Politecnico di Torino (Italy)
Abstract: A geometrical structure is provided for classes of high-dimensional Bernoulli distributions that turn out to be important to address open issues in the study of their statistical properties, such as dependence or aggregate risk. The analyzed classes are convex polytopes; in some cases, the extremal generators can be analytically provided; in some other cases, their extremal generators are a more challenging task that can be addressed using an algebraic representation. The class of multivariate Bernoulli distributions are considered to have identical marginal Bernoulli distributions with mean p and the class of multivariate Bernoulli distributions with given sums. The results are applied to study lower bounds for the risk of a credit portfolio.
