Abstract
A new method for estimating the power spectral density functions of stationary autoregressive-moving average processes is proposed. In this method, the psdf is obtained only from the estimated AR parameters and sample auto-correlations of the processes. MA parameters are not needed for the estimation of psdf. AR parameters are estimated so as to minimize the sum of squared residuals of Yule-Walker's equations. A new criterion for estimating the order of AR part of the process is introduced. It is shown that the estimate of psdf by this method is asymptotically unbiased. Numerical experiments show that this method gives us good estimates.
