Construction of bootstrap confidence intervals on sensitivity indices computed by polynomial chaos expansion

Construction of bootstrap confidence intervals on sensitivity indices computed by polynomial chaos expansion

0.00 Avg rating—0 Votes
Article ID: iaor20141399
Volume: 121
Issue: 3
Start Page Number: 263
End Page Number: 275
Publication Date: Jan 2014
Journal: Reliability Engineering and System Safety
Authors: , , ,
Keywords: chaos, metamodels, sensitivity analysis, Statistics: bootstrap, confidence interval
Abstract:

  • The proposed methodology combines advantages of sparse polynomial chaos expansion with bootstrap re‐sampling to compute variance‐based sensitivity indices.
  • A conservative way to choose the number of bootstrap re‐sampling is presented.
  • A method to increase the degree of the polynomial basis, linked to the size of confidence intervals, is proposed.
  • Comparisons with classical meta‐model error estimators reveals the interest of a sensitivity‐indices‐oriented methodology.

    • Reviews

    Required fields are marked *. Your email address will not be published.