The sequential fitting of subset auto regressions with a forgetting factor

In this paper, we propose two forward time update algorithms for the covariance case to recursively estimate subset autoregressive models with a forgetting factor. These time update algorithms in conjunction with the forgetting factor approach apply appropriate weights to update the model structure and parameters. With forgetting factor and time update algorithms we are able to properly analyse complex relationships where the relevant economic series have been generated from structures subject to evolutionary change in their environment. An illustration of these procedures is presented using the spot aluminium and nickel prices on the London Metal Exchange. The proposed algorithms are also applicable to subset AR modelling without a forgetting factor.