Assessing the association between two spatial or temporal sequences

Abstract

This paper deals with the codispersion coefficient for spatial and temporal series. We present some results and simulations concerning the codispersion coefficient in the context of spatial models. The results obtained are immediate consequences of the asymptotic normality of the sample codispersion coefficient and show certain limitations of the coefficient. New simulation studies provide information about the performance of the coefficient with respect to other coefficients of spatial association. The behavior of the codispersion coefficient under additively contaminated processes is also studied via Monte Carlo simulations. In the context of time series, explicit expressions for the asymptotic variance of the sample version of the coefficient are given for autoregressive and moving average processes. Resampling methods are used to compute the variance of the coefficient. A real data example is presented to explore how well the codispersion coefficient captures the comovement between two time series in practice.

Keywords:

• spatial association
• autoregressive models
• correlation coefficient
• codispersion coefficient
• time series

Acknowledgements

The author is grateful to Manuel Galea and Victor Leiva for their helpful comments and suggestions. The author also thanks two anonymous referees and the editor for their valuable suggestions. This research was supported by Fondecyt Grant 11075095, Chile.