Abstract. A formal justification for the use of the method of autoregressive spectral estimation for time series consisting of a sinusoidal signal in additive noise is given in this paper. The analytical properties of the autoregressive approximation to the generalized spectral density of the process are presented, and the operational characteristics of the statistical estimation procedure are discussed. In particular, strong convergence of the autoregressive parameters and the autoregressive transfer function approximation is shown.
|Number of pages||13|
|Journal||Journal of Time Series Analysis|
|Publication status||Published - 1 Jan 1990|
- mixed spectra
- transfer function