A Practical and Fast Numerical Method for Calculating Global Sensitivity with Examples from Supply Chain and Measurement Applications

Global sensitivity analysis (GSA) is an essential tool for understanding, verifying, and improving mathematical models of stochastic systems. In GSA the influence of a model’s output uncertainty is apportioned among the model’s inputs and, the influence of each input is ranked with a sensitivity index. The Sobol sensitivity indices are cited frequently in the literature due to their effectiveness and sound theoretical basis. It is common practice to calculate the Sobol first-order and total effect indices (called the First_Total method). Following the design science methodology for information science, a new method, called First_CMD, is presented which provides a sensitivity ranking similar to the First_Total, but can be computed in orders of magnitude less time. A review of supply chain models shows that interest in uncertainty analysis is growing, especially for green (sustainable) supply chains. However, GSA is rarely mentioned in supply chain literature. It is shown how GSA, using First_Total or First_CMD can be applied to an existing model which determines sustainability metrics for international biodiesel supply chains. The computational time and rankings are compared for GSAs carried out using both First_Total and First_CMD for a complex model developed by Keysight Technologies. The model contains 176 stochastic inputs and simulates the uncertainty of measurements taken by an electronic system. It is shown that First_CMD yields similar information as First_Total, but the computation time of uncertainty and sensitivity analysis is reduced from approximately one year to one day due to fewer model evaluations