Computational aspects in checking of coherence and propagation of conditional probability bounds

In this paper we consider the problem of reducing the computationaldifficulties in g-coherence checking and propagation of imprecise conditionalprobability assessments. We review some theoretical results related with thelinear structure of the random gain in the betting criterion. Then, we proposea modi ed version of two existing algorithms, used for g-coherence checkingand propagation, which are based on linear systems with a reduced number ofunknowns. The reduction in the number of unknowns is obtained by an iterativealgorithm. Finally, to illustrate our procedure we give some applications.