Computational practice: Multivariate parametric or nonparametric modelling of european bond volatility spillover?
Date of Original Version
Previous research has documented the effectiveness of multivariate nonparametric radial basis function artificial neural networks to model the simultaneous bi-directional volatility spillover effects among European economies. As a nonparametric estimation method, artificial neural networks do not present researchers with the same confidence levels on weight estimation that are commonplace under the assumption of asymptotic normality under linear regression. This chapter considers extending prior research findings by examining the domain of applicability for linear multivariate parametric model when applied to the estimation of global government bond volatility spillover models. To this end the chapter examines both multivariate linear regression and canonical correlation techniques to establish a comparative set of findings to those presented from prior research using a multivariate radial basis function artificial neural network. The findings clearly demonstrate that linear parametric methods fail to adequately explain the correlation and cross-correlation structure of excess European bond returns. Further, for studies designed to map the continuity of cross-border bond volatility spillover, the research demonstrates the overall effectiveness of neural networks to map such real-valued measurable functions. © 2013 by Nova Science Publishers, Inc. All rights reserved.
Recent Advances in Computational Finance
Kajiji, Nina, and Gordon H. Dash. "Computational practice: Multivariate parametric or nonparametric modelling of european bond volatility spillover?." Recent Advances in Computational Finance , (2013): 187-204. https://digitalcommons.uri.edu/cba_facpubs/220