Benchmark of Goal Oriented Sensitivity Analysis Methods Using Ishigami Function

Abstract: The present paper deals with the Goal Oriented Sensitivity Analysis. The median-oriented sensitivity analysis (MSA) is presented which measures sensitivities by applying absolute distances between model outputs Y and median of Y. The median is used as the alternative central parameter to arithmetical mean applied by the established Sobol sensitivity analysis (SSA). General agreements and differences between MSA and SSA are studied by applying the Ishigami function. The paper shows that the sensitivity analysis need not necessarily be based on the analysis of variance known as ANOVA, but that there exist alternate approaches, too. CPU demanding character of MSA is approximately identical to that of SSA. The proposed MSA is efficient and practical for the problems in which it is necessary to quantify the importance of each input variable with respect to the median.