Abstract
Various ways of approximation of the true probability distribution by a discrete distribution concentrated in a finite number of atoms, called scenarios, are classified. Resistance of the numerical results with respect to the choice of scenarios and/or of their probabilities and also the relation between results obtained for the selected scenarios and those valid for the true underlying distribution are of great practical interest.
Various approaches to output analysis in the framework of an expectation type stochastic program are discussed, bearing in mind the diverse origin of scenarios.