Sensitivity Analysis of Model Output with Input Constraints: A Generalized Rationale for Local Methods

Abstract

In this work, we introduce a generalized rationale for local sensitivity analysis (SA) methods that allows to solve the problems connected with input constraints. Several models in use in the risk analysis field are characterized by the presence of deterministic relationships among the input parameters. However, SA issues related to the presence of constraints have been mainly dealt with in a heuristic fashion. We start with a systematic analysis of the effects of constraints. The findings can be summarized in the following three effects. (i) Constraints make it impossible to vary one parameter while keeping all others fixed. (ii) The model output becomes insensitive to a parameter if a constraint is solved for that parameter. (iii) Sensitivity analysis results depend on which parameter is selected as dependent. The explanation of these effects is found by proposing a result that leads to a natural extension of the local SA rationale introduced in Helton (1993). We then extend the definitions of the Birnbaum, criticality, and the differential importance measures to the constrained case. In addition, a procedure is introduced that allows to obtain constrained sensitivity results at the same cost as in the absence of constraints. The application to a nonbinary event tree concludes the article, providing a numerical illustration of the above findings.

Variable importance analysis: A comprehensive review

“…Other local VIMs include the differential importance measure (DIM) [50][51][52] and the finite change decomposition based VIMs [53], both of which attribute the finite change of model output to each of the input and/or the interactions of inputs with different strategies so as to measure the individual effect of the small change of each input and their interaction effects on the change of model output. One can refer to the respective reference for details.…”

Sensitivity Analysis Using Risk Measures

“…Straightforward calculus leads to: Therefore, sensitivity to the group G of model inputs decomposes into a sum of single‐factor sensitivities. This form of additivity is shared with the (local) differential importance measure , which calculates sensitivity as the ratio of a partial derivative over the total differential. In this respect, the sensitivity method proposed here can be seen as an extension of the differential importance measure, applying to functionals.…”

Sensitivity Analysis of Environmental Models: A Systematic Review with Practical Workflow

“…(1) is not differentiable at x. Derivative-based SA finds its rationale in the Taylor series expansion. This is well explained in Helton (1993) and generalized later on in Borgonovo (2008). In order to facilitate a comparison of sensitivities across input factors that may have different units of measurements, the partial derivatives are usually rescaled (e.g.…”