Multivariate AR Model Order Estimation with Unknown Process Order.

Autor: Bozanis, Panayiotis, Houstis, Elias N., Pappas, Stylianos Sp., Leros, Assimakis K., Katsikas, Sokratis K.
Zdroj: Advances in Informatics; 2005, p787-797, 11p
Abstrakt: A new method for simultaneous order estimation and parameter identification of a multivariate autoregressive (AR) model is described in this paper. The proposed method is based on the well known multimodel partitioning theory. Computer simulations indicate that the method is 100% successful in selecting the correct model order in very few steps. The results are compared with another two established order selection criteria the Akaike's Final Prediction Error (FPE) and Schwarz's Bayesian Criterion (BIC). [ABSTRACT FROM AUTHOR]
Databáze: Supplemental Index