Affordable Access

Publisher Website

Screening important inputs in models with strong interaction properties

Authors
Publisher
Elsevier Ltd
Publication Date
Volume
94
Issue
7
Identifiers
DOI: 10.1016/j.ress.2008.10.007
Keywords
  • Screening Design
  • Elementary Effect Method
  • Factorial Design
  • Variance-Based Sensitivity Measures
Disciplines
  • Computer Science
  • Design
  • Mathematics

Abstract

Abstract We introduce a new method for screening inputs in mathematical or computational models with large numbers of inputs. The method proposed here represents an improvement over the best available practice for this setting when dealing with models having strong interaction effects. When the sample size is sufficiently high the same design can also be used to obtain accurate quantitative estimates of the variance-based sensitivity measures: the same simulations can be used to obtain estimates of the variance-based measures according to the Sobol’ and the Jansen formulas. Results demonstrate that Sobol’ is more efficient for the computation of the first-order indices, while Jansen performs better for the computation of the total indices.

There are no comments yet on this publication. Be the first to share your thoughts.