Asymptotic accuracy of identifying a discrete plant by a generalized least squares method

An asymptotic formula is obtained for the dependence of the mean squore error in forecasting the response of a linear plant whose parameters are computed by a generalized least squares method (GLSM) on statistical characteristics of the input signal and noise provided that these are auto-regression firts-order processes. The sensitivity of this error to variations in parameters of the GLSM weighting matrix is assesed. Accuracies obtainable in the GLSM and ordinary LSM are compared within the framework of this problem.