A New computational method of a moment-independent uncertainty importance measure

Liu, Q.; Homma, Toshimitsu

The uncertainty of input parameters are transferred through the model to the output and leads to the issue of output uncertainty. The study of how the uncertainty in the model output can be apportioned to the uncertainty in the inputs is the job of sensitivity analysis. In this paper, the authors analyzed a newly proposed uncertainty importance measure, Delta. The core calculation method of Delta is based on the measurement of the area between the conditional PDF (Probability Density Functions) and the unconditional PDF of the model output. The authors found that the area between these two PDFs is equivalent to two times of the algebraic sum of the vertical deviations between their corresponding CDFs at the intersection points of these two PDFs. Therefore, the author proposed a new method to calculate Delta. The applicability of this new method is proved by applying it to two test models.