Dear Professor,
I use Sobol indices based on PCE to do sensitivity analysis about a complex structure. There are 20 random input variables. The final LOO error estimate is about 0.18, which is minimized in all tried ExpDesign.
However, the sum of all the Sobol indices of random input variables is more than 1. Is it right? If it is right, how to explain it?
Best wishes!
Tony Smith