Efficient global sensitivity analysis of computer simulation models using an adaptive Least Angle Regression scheme

An efficient method is proposed to perform global sensitivity analysis of a the computer model of a physical system whose input parameters are random. The so-called sensitivity indices are usually computed by Monte Carlo simulation, which reveals inappropriate for industrial applications due to an unaffordable computational cost. In the present paper, polynomial chaos (PC) expansions are introduced in order to compute sensitivity indices. An adaptive Least Angle Regression (LAR) algorithm is devised for automatically detecting the significant coefficients of the PC expansion. The latter may thus be computed eventually by means of a relatively small number of possibly costly model evaluations. The method is applied to the sensitivity analysis of the so-called Sobol’ function.