Modeling Influential Correlated Noise Sources in Multivariate Dynamic Systems

Y. Tanokura and G. Kitagawa (Japan)


Model Development, Detection of Noise Source, Akaike's Power Contribution, Multivariate Time Series


Akaike's power contribution, a useful concept in detecting noise sources of multivariate dynamic systems with feed back, is not applicable to the systems with high correlations of noise due to the assumption of independence of noise. To address this problem, we develop a decomposition of a general variance covariance matrix of noise, modeling cor relations among variables, and define a new power contri bution that extends Akaike's concept. It was shown that the extended power contribution succeeds in modeling the in dependent and correlated noises and that Akaike's original power contribution precisely captures the former. In the ex ample of the US macroeconomic data, the newly obtained information on correlated noises was presented.

Important Links:

Go Back