ARMA modeling based on partial AR and MA parameter approximation

A transfer-function modeling concept is developed for an unknown casual stable, linear-time invariant system. The approach taken has been to treat the models separately and then to achieve the general autoregressive-moving-average (ARMA) transfer function representation. The AR approach has been obtained from conventional methods of the Yule-Walker formulations and associated Levinson-Durbin algorithm. A somewhat different approach based on the Tauberian approximation is proposed to achieve the MA modeling.<
>