An asymptotically efficient autoregressive moving-average (ARMA) spectral estimator is presented, based on the sample covariances of observed time series. The estimate of the autoregressive (AR) part is shown to be identical to the optimal instrumental variable (IV) estimator in  although derived here using a different approach. The moving-average (MA) spectral parameter estimate is new.
|Number of pages||4|
|Journal||IEEE Transactions on Automatic Control|
|State||Published - Nov 1986|