Comparison of two sets of Monte Carlo estimators of Sobol’ indices

This study compares the performances of two sampling-based strategies for simultaneous estimation first- and total-order variance-based sensitivity indices (a.k.a. Sobol’ indices). The first strategy corresponds to current approach employed by practitioners recommended in literature. second one was only recently introduced last authors present article. Both rely on different estimators indices. asymptotic normal variances sets are established their accuracies compared theoretically numerically. results show that new outperforms one. global analysis radiative forcing model sulfur aerosols is performed with strategy. confirm this interactions important input variable irrelevant.