The EASI algorithm is a Fourier-based technique for performing variance-based methods of global sensitivity analysis for the computation of first order effects (a.k.a. Sobol’ indices, main effects, correlation ratios), hence belonging into the same class of algorithms as FAST and RBD. Algorithms of this type are using a frequency-based approach, i.e., signals of known frequencies are assigned to the input factors, and a frequency analysis is carried out on the output that computes the influence of each input factor on the output, see Fig. 5 for a demonstration of a frequency response.

EASI is a computationally cheap method for which existing data can be used. Unlike the FAST and RBD methods which use a specially generated sample set that contains suitable frequency data for the input factors, in EASI these frequencies are introduced by sorting and shuffling the available input samples. The sorted input will be a nearly symmetric, periodic signal of frequency 1 (irrespectively of the input distribution). The output data are sorted accordingly matching the resorted input samples, hence avoiding re-evaluation of the model. These sorted data can then be analysed using the power spectrum of the output. This latter analysis forms the standard back-end procedure of the Fourier-based techniques.

For higher-order effects, the sorting algorithm is implemented via a multi-dimensional search curve. The sorted and shuffled input obtained with this method will be a perturbed signal with a certain frequency spectrum. It is hoped that the non-periodic perturbations are distributed over the whole spectrum and therefore are of little influence for the spectral analysis.

• Elmar Plischke, An effective algorithm for computing global sensitivity indices (EASI), Reliability Engineering & System Safety, Volume 95, Issue 4, April 2010, Pages 354-360.