Study of conditional ML estimators in time and frequency-domain system identification (Q1295116)

Linear dynamic systems are considered in the time domain and in the frequency domain. It is shown that the full covariance matrix can be replaced by its main diagonal in maximum likelihood estimation, and both respective estimates have an identical asymptotical behaviour. The result is very important because it provides a theoretical justification for the common practise to neglect the initial condition effects of the noise model.