Quantitative methods for global sensitivity analysis
This paper presents an extension of the standard Fourier amplitude sensitivity test (FAST) which aims at the computation of total sensitivity measures of model output(s). A total sensitivity measure/index takes into account a parameter's main effect (that FAST is capable of computing up until now) as well as all the interactions of any order involving that parameter. Sobol' measure is another technique that also offers the computation of total sensitivity indices for non-linear and non-monotonic models. This paper presents results of two simulation studies which show the extended FAST is not only and alternative technique to Sobol' but it is also computationally more efficient, the estimates have the same precision as the Sobol's, especially at low sample size. Results of one of the simulation studies also demonstrate that the incorporation of total effect indices is the only way to perform a rigorous quantitative sensitivity analysis.
Bibliographic Reference: Paper presented: ESREL '97, June 17-20, 1997
Availability: Available from (1) as Paper EN 40693 ORA
Record Number: 199710904 / Last updated on: 1997-07-23
Original language: en
Available languages: en