Calibrating agent-based models with linear regressions

Abstract

In this paper, we introduce a simple way to parametrize simulation models by using regularized linear regression. Regressions bypass the three major challenges of calibrating by minimization: selecting the summary statistics, defining the distance function and minimizing it numerically. By substituting regression with classification, we can extend this approach to model selection. We present five example estimations: a statistical fit, a biological individual-based model, a simple real business cycle model, a non-linear biological simulation and heuristics selection in a fishery agent-based model. The outcome is a method that automatically chooses summary statistics, weighs them and uses them to parametrize models without running any direct minimization.