Stochastic Dominance Constraints induced by Mixed-Integer Linear Recourse

Abstract

As described in Section 1.1.1, solving a two-stage stochastic optimization problem corresponds to the selection of a random variable. Applying a traditional meanrisk approach to this problem, requires to select the random variable with respect to statistical parameters reflecting mean and/or risk, like the expectation, the random variable’s excess over some preselected ruin level, or the variable’s quantiles. In all cases, a “best” random variable is chosen among all possible random variables. Only the definition of “best” distinguishes the different mean-risk formulations.