|
Abstract:
|
We consider the problem of sensitivity analysis in computer experiments, in which the user of a computer model wishes to identify the most important uncertain inputs in the model. Decision-theoretic arguments based on the expected value of perfect information for are used to quantify the importance of each uncertain input. The sensitivity analysis can then guide the model user in deciding whether it would be worthwhile to first collect more data about the uncertain inputs, before making decisions based on the model outputs. Standard Gaussian process emulators are used for efficient computation, and we address the problem of quantifying uncertainty in the sensitivity analysis results due to the use of an emulator with limited model runs.
|
