Bounds for functions of multivariate risks (Q2489767)

The authors consider the problem to derive bounds for the probability that \(\Psi (X_1, \ldots, X_n)\) \(< s\), where \(X_i\) are any \(k\)-dimensional random vectors with df \(F_i, s \in \mathbb{R}^k\), and \(\Psi\) is componentwise increasing. The bounds are derived as consequences of the duality theorem for the marginal problem. Results of this type figure in the literature as Hoeffding-Fréchet bounds. For the case of the sum functional they derive an improvement and correction of corresponding results by \textit{H. Li, M. Scarsini} and \textit{M. Shaked} [J. Multivariate Anal. 56, No. 1, 20--41 (1996; Zbl 0863.62049)]. The case of multivariate uniform random vectors can be solved explicitly.