Adaptive use of replicated Latin Hypercube Designs for computing Sobol’ sensitivity indices

As recently pointed out in the field of Global Sensitivity Analysis (GSA) computer simulations, use replicated Latin Hypercube Designs (rLHDs) is a cost-saving alternative to regular Monte Carlo sampling estimate first-order Sobol’ indices. Indeed, two rLHDs are sufficient compute whole set those indices regardless number input variables. This relies on permutation trick which, however, only works within class estimators called Oracle 2. In present paper, we show that still beneficial another estimators, 1, which often outperforms 2 for estimating small and moderate Even though unlike computation cost 1 depends dimension, can be applied construct an averaged (triple) estimator whose great accuracy presented numerical example. Thus, promote adaptive rLHDs-based sensitivity analysis where first stage by If needed, then reevaluated estimators. strategy, guaranteeing estimates, model from nuclear field.