A fundamental relation between predictor identification and power spectrum estimation

SummaryAn asymptotic linear relation between the FPE (final prediction error) of the predictor obtained by using the least squares estimate of autoreggression coefficients and the integrated relative mean square error of the estimate of power spectrum obtained from the fitted autoregression model is established. This relation provides a link between the predictor identification and the power spectrum estimation and has already been used as a theoretical background of a former paper [3].