论文检索

时间序列模型频域估计中的偏差校正AI 解读

Exact expressions for the expectations and covariances of frequency domain estimators for integrals of the spectrum of a stationary time series are derived. It is demonstrated that the obvious correction for bias improves the estimator even when observations are missing. The techniques are applied to frequency domain estimation of model parameters, and this is illustrated by an application on a medical time series with missing observations.

时间序列模型的拟合优度检验AI 解读

A goodness-of-fit test statistic for time series models is presented, and its asymptotic distribution, even when parameters are estimated, is derived. The statistic is a frequency domain analogue of the Box-Pierce portmanteau statistic. The asymptotic power of the test is found and compared to the empirical power of the portmanteau test. The small sample properties are investigated by Monte-Carlo methods.