Efficient Bias Corrected Nonparametric Spectral Estimation
针对特定频率的谱密度估计，基于核谱估计的局部二阶行为建立线性模型，推导出第二阶段估计量，在平滑参数过度平滑时渐近更优，模拟和实例验证了其良好性能。
We consider estimation of a spectral density at a particular frequency. A linear model is developed based on the local second order behaviour of kernel spectral estimates and a second stage estimate is derived from this model. The estimate is shown to be asymptotically more efficient than the kernel estimate with the optimal, but unknown, bandwidth whenever the smoothing parameter is chosen so that it oversmooths the kernel estimate. A small-sample simulation study shows good characteristics of the second stage estimate. Practical implementation is discussed through examples.