CFE 2019: Start Registration
View Submission - CFE
A1924
Title: Model calibration and validation via confidence sets Authors:  Raffaello Seri - University of Insubria (Italy) [presenting]
Mario Martinoli - University of Insubria, Sant'Anna School of Advanced Studies (Italy)
Samuele Centorrino - Stony Brook University (United States)
Davide Secchi - University of Southern Denmark (Denmark)
Abstract: An earlier calibration and validation method for simulation models is extended. The previous method was based on the concept of Model Confidence Set, as is the present one. Given a distance between time series, a benchmark dataset, and a finite set of simulation models $\mathcal{M}$, the method allowed the researcher to build a confidence set, obtained as a subset of $\mathcal{M}$, containing, with prescribed probability, the model (or models) minimizing the distance with respect to the data. The main drawback of the method was that, in accordance with most approaches to calibration, it neglected the variability of the data and focused on the simulations. We investigate the effects of the variability of the data on the procedure and, when necessary, we propose some modifications.