A Novel Sensitivity Analysis for Dynamic Models

摘要

This paper proposes a novel statistical sensitivity analysis for dynamic models, which is based on an enhanced maximal information coefficient (MIC) method. We enhanced the MIC method to handle multivariate sensitivity analysis; rather than just univariate analysis. The main motivation of this enhancement is overcoming the research gaps in the current state of the art of eigenvalue analysis. We postulate that this novel sensitivity analysis represents a solid foundation to study the multivariate complex nonlinear non-monotonic relationships between behavior modes - expressed by eigenvalues - and the model parameters. The experiments conducted corroborate our postulation.