Sensitivity analysis for chemical models

The convergence of Fourier amplitude sensitivity test (FAST)-based and sampling-based strategies for sensitivity analysis and also a convergence between these and experimental design theory vindicate the original intuition of the FAST developers that Si (model free extension) was a good measure for chemical models. Both FAST and MC computation schemes have been upgraded in recent years, becoming less expensive and easier to apply. Nevertheless, one-factor-at-a-time (OAT) approaches still dominate the chemical literature even when the purpose of the analysis is to assess the relative importance of input factors in the presence of factors uncertainty. To contrast this practice, this review covers the most promising modern approaches to sensitivity analysis, with some emphasis on the methods in the FAST family, comparing, with the help of worked examples, the performances of variance-based methods with different kinds of local, regression, or screening-based measures.