5. Model sensitivity analysis

5.1 Principle

Let's consider a mathematical model, made up of a set of random input variables X = (X ₁ ,..., X _p ), a deterministic function f and a random output (response) variable Y. This function can be very complex, and is in practice evaluated using computer code, which is more or less computationally expensive. Sensitivity analysis studies how, and above all how intensely, disturbances to the model inputs cause disturbances to the response. We're only interested here in global sensitivity analysis, which studies how input variability affects response variability, by determining how much of the variance in response is due to a given input or set of inputs. It also highlights the input variables...