A new algorithm for the ARMA model estimation

The authors study the full structure estimation of the ARMA (autoregressive moving average) time series. First, the AR-part´s order p is discriminated by a determinant ratio criterion. Second, the authors propose a fast algorithm of the order recursive overdetermined modified Yule-Walker estimation method in which the parameters and the order of the AR-part of the ARMA model are estimated simultaneously. Finally, the MA-part´s parameters and order of the ARMA model are given using the estimated purely AR model parameters as data. The estimation of the ARMA model is thus completed