Abstract
We show that variance-based, density-based, expected-value of information based sensitivity measures rest on the same rationale, namely, information updating. We prove general results concerning their properties. We then show that they can all be estimated by the same design. The design is suitable in the presence of correlated inputs and makes the estimation cost independent of the number of model inputs and depending solely on the sample size. Results of numerical experiments are proposed.
| Original language | English (US) |
|---|---|
| Pages | 3255-3260 |
| Number of pages | 6 |
| State | Published - Jan 1 2014 |
| Event | European Safety and Reliability Conference, ESREL 2013 - Amsterdam, Netherlands Duration: Sep 29 2013 → Oct 2 2013 |
Other
| Other | European Safety and Reliability Conference, ESREL 2013 |
|---|---|
| Country | Netherlands |
| City | Amsterdam |
| Period | 9/29/13 → 10/2/13 |
Fingerprint
ASJC Scopus subject areas
- Safety, Risk, Reliability and Quality
Cite this
}
A common rationale for global sensitivity analysis. / Borgonovo, E.; Hazen, Gordon B; Plischke, E.
2014. 3255-3260 Paper presented at European Safety and Reliability Conference, ESREL 2013, Amsterdam, Netherlands.Research output: Contribution to conference › Paper
TY - CONF
T1 - A common rationale for global sensitivity analysis
AU - Borgonovo, E.
AU - Hazen, Gordon B
AU - Plischke, E.
PY - 2014/1/1
Y1 - 2014/1/1
N2 - We show that variance-based, density-based, expected-value of information based sensitivity measures rest on the same rationale, namely, information updating. We prove general results concerning their properties. We then show that they can all be estimated by the same design. The design is suitable in the presence of correlated inputs and makes the estimation cost independent of the number of model inputs and depending solely on the sample size. Results of numerical experiments are proposed.
AB - We show that variance-based, density-based, expected-value of information based sensitivity measures rest on the same rationale, namely, information updating. We prove general results concerning their properties. We then show that they can all be estimated by the same design. The design is suitable in the presence of correlated inputs and makes the estimation cost independent of the number of model inputs and depending solely on the sample size. Results of numerical experiments are proposed.
UR - http://www.scopus.com/inward/record.url?scp=84900026645&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84900026645&partnerID=8YFLogxK
M3 - Paper
SP - 3255
EP - 3260
ER -