Project CEMAPRE internal
|Title||Screening and Confounding|
|Participants||Pierre Barbillon, Rui Paulo (Principal Investigator)|
|Summary||When confronting simulators with experimental data, a common approach involves acknowledging a|
source of uncertainty known as model discrepancy. This is the materialization of the idea that any
given mathematical model, no matter how sophisticated, will not be able to perfectly reproduce a
real phenomenon. The experimental data is hence modeled as the sum of three components:
observational error, simulator at the "best" value of the vector of model parameters, and the bias
function. There is a considerable literature on the topic of confounding between these components.
In particular, it is well known that the bias function and the vector of model parameters may be
seriously confounded. Recently, we have proposed methodology to screen the bias function for active
inputs and simultaneously estimate the vector of model parameters. We are now interested in
investigating the effect of the confounding in the ability to effectively identify the active inputs
in the bias function.