Dear Uqlab team,
Here, I have a question to ask, how do I get an explicit function of input variables and responses when I perform Kriging metamodeling using existing data? Because an explicit function is required in the subsequent sensitivity analysis.
For example, in the UQlab case Kriging from existing data1: truss, how to perform sensitivity analysis after metamodeling? Is it convenient to introduce the specific steps in detail?
Thank you in advance!
Dear FANFAN,
In principle your new model will be the Kriging metamodel and you can perform the sensitivity based on it directly. Also you can simply set Mymodel = myKriging; and in this case you can proceed with the sensitivity analysis. A clear steps can be found in the example of Sobol indices, where you can use the Kriging metamodel the same as PCE is used in this example:
or see this
There was a comment by Prof. Sudret regarding sensitivity based on Kriging ** make sure that the surrogate modelling error is small enough (e.g., leave-one-out error smaller than 0.01).**
Hope this answers your question!