Loss of Spectral Peaks in Autoregressive Spectral Estimation
通过渐近分析发现，Yule-Walker估计会导致谱峰丢失和严重偏差，而最小二乘估计更优，模拟结果也支持这一结论。
Autoregressive spectral analysis depends on the method used for estimating the autoregressive parameters. It is shown by an asymptotic analysis involving second-order terms that least-squares estimates should be preferred to Yule-Walker estimates, since Yule-Walker estimates may result in loss of spectral peaks and strong bias. The results are confirmed by simulations which include also Burg-type estimates.