Résumé
Variance-based sensitivity analysis and multivariate sensitivity analysis aim to apportion the variability of the model output(s) into input factors and their interactions. Sobol's total index, which accounts for the effects of interactions, serves as a practical tool to deal with the curse of dimensionality. In this paper, we address the problem of the efficient estimation of Sobol's total index. First, we provide a generalized and optimal estimator of the variance of the total effect function, including Jansen's estimator, its rate of convergence, and its asymptotic distribution; second, we derive the asymptotic distribution of total indices; and third, we investigate the applicability of these results to allow for improving the estimation of the total indices for some specific degrees of the kernel.