ARMA、结构时间序列、动态回归和缺失观测可放进统一状态表示;这种表示等价需给出具体最小性、初始化与噪声假设,因此本文不在 metadata 中笼统声明 state-space 与 ARMA 等价。初始化尤其关键:稳定状态可用平稳协方差,随机游走水平需弥散或显式先验。错误地用固定小方差初始化,会让早期滤波过度自信并扭曲似然。
参考资料
James Durbin and Siem Jan Koopman, Time Series Analysis by State Space Methods, 2nd ed., Oxford University Press, 2012,Chs. 2–3,state-space formulation and linear Gaussian models。
Robert H. Shumway and David S. Stoffer, Time Series Analysis and Its Applications, 4th ed., Springer, 2017,Ch. 6,state-space models and dynamic linear models。
Andrew C. Harvey, Forecasting, Structural Time Series Models and the Kalman Filter, Cambridge University Press, 1989,Chs. 2–3,structural state-space modeling。