Stochastic spectral embedding (SSE) for Global sensitivity analysis

UQ with UQLab Community Q&A and How To
1

Hi, Can stochastic spectral embedding (SSE) be used for global sensitivity analysis?
Then how to implement this?
Could you give an example?

2

Hi @AmosJoseph,

Thanks for posting here. In fact, you can simply use SSE as a standard UQLab model and apply Monte Carlo simulation for global sensitivity analysis. For example:

SobolSensOpts.Type = 'Sensitivity';
SobolSensOpts.Model = MySSE;
SobolSensOpts.Input = myInput;
SobolSensOpts.Method = 'Sobol';
SobolSensOpts.Sobol.Sampling = 'LHS';
SobolSensOpts.Sobol.SampleSize = 1e5;
SobolAnalysis = uq_createAnalysis(SobolSensOpts);

where we just need to provide our SSE model to the field “Model” of the sensitivity option (i.e., the second line of the code above). The other settings are classical options for global sensitivity analysis, which you are free to modify according to the concrete case study (please refer to the sensitivity manual for more details).

I hope this helps.

Xujia

1 Like
3

Thanks a lot for your help！