Evidence-invariant sensitivity bounds
Silja Renooij, Linda C. van der Gaag · 2004
The sensitivities revealed by a sensitivity anal-ysis of a probabilistic network typically depend on the entered evidence. For a real-life network therefore, the analysis is performed a number of times, with different evidence. Although efficient algorithms for sensitivity analysis exist, a com-plete analysis is often infeasible because of the large range of possible combinations of obser-vations. In this paper we present a method for studying sensitivities that are invariant to the ev-idence entered. Our method builds upon the idea of establishing bounds between which a parame-ter can be varied without ever inducing a change in the most likely value of a variable of interest. 1