Research projects

Project CEMAPRE internal

ParticipantsPierre Barbillon, Gonzalo Garcia-Donato, Rui Paulo (Principal Investigator)
SummaryMathematical models (or computer models, or simulators) designed to simulate real complex phenomena
are increasingly being constructed and used in all fields of science.

Computer models can be seen as machines that transform inputs, variables describing the
configuration of the system, into outputs, the computer-simulated value of the phenomenon that we
are modelling. The purpose of this project is to further explore and compare Bayesian statistical
methodology to ascertain which inputs most significantly affect the output, the co-called active
inputs, an activity commonly known as screening.

The simulator is almost always a simplified description of reality, and the discrepancy between the
simulated value and reality is often modelled as a function of the inputs, through the so-called
bias function. An additional, and quite innovative, topic of research in this project is to explore
the possibilities of using the aforementioned Bayesian statistical methodologies to detect active
inputs in the bias function, effectively determining in which inputs the simulator is failing to
capture reality.