An alternative way to compute Fourier amplitude sensitivity test (FAST)

In this article we investigate the relationship between two coefficients used in sensitivity analysis of model output. One is the Fourier amplitude sensitivity test's coefficient, developed in the 1970s, and the other the Sobol' sensitivity indices, developed in the 1990s. Supposedly both methods are capable of computing the "main effect" contribution of model's input parameters to model's output variance. We discuss the equivalence of the two methods, and prove the identity of their prediction on two test cases. Relative advantage and disadvantages of the methods are also illustrated. (C) 1998 Elsevier Science B.V. All rights reserved.