The first output must be the first order sensitivity indices unbiased.
The biased version of sensitivity indices are wrong, i think you can make them available but in a debug mode only. To replace them, you can add the partial variance ( si_c * v )
In the Bootstrap part, the mean value of SI's is not very informative and it is a worst indicator than the initial sensitivity indice (calculated with the full set). You can replace it with the 5th and 95th percentile witch represent the limits values
Hi !
The first output must be the first order sensitivity indices unbiased. The biased version of sensitivity indices are wrong, i think you can make them available but in a debug mode only. To replace them, you can add the partial variance ( si_c * v )
In the Bootstrap part, the mean value of SI's is not very informative and it is a worst indicator than the initial sensitivity indice (calculated with the full set). You can replace it with the 5th and 95th percentile witch represent the limits values
Best.