Community Research and Development Information Service - CORDIS

Abstract

A number of methods are available to assess uncertainty importance in the predictions of a simulation model for orthogonal sets of uncertain input factors. However, in many practical cases input factors are correlated. Even for these cases it is still possible to compute the correlation ratio and the partial (or incremental) importance measure, two popular sensitivity measures proposed in the recent literature on the subject.

However, the existing indicators of importance have limitations in terms of their use in sensitivity analysis of model output. Correlation ratios are indeed effective for priority setting (i.e. to find out what input factor needs better determination) but not, for instance, for the identification of the subset of the most important input factors, or for model simplification. In such cases other types of indicators are required that can cope with the simultaneous occurrence of correlation and interaction (a property of the model) among the input factors. .

Additional information

Authors: SALTELLI A, JRC, ISIS, Ispra (IT);RATTO M, JRC, ISIS, Ispra (IT);TARANTOLA S, JRC, ISIS, Ispra (IT)
Bibliographic Reference: An oral report given at: European Safety and Reliability International Conference. Held in: Torino (IT), 16-20 September 2001
Record Number: 200113909 / Last updated on: 2001-10-22
Category: PUBLICATION
Original language: en
Available languages: en
Follow us on: RSS Facebook Twitter YouTube Managed by the EU Publications Office Top