No label defined (Q4217907)

General state space modeling, initiated by \textit{G. Kitagawa} [J. Am. Stat. Assoc. 82, 1032-1063 (1987; Zbl 0644.62088)], is a Bayesian modeling of not-necessarily linear not-necessarily Gaussian time series. It is an extension of our earlier work on ``smoothness priors'', a Bayesian linear Gaussian modeling of times series. In Section II, background material including Akaike's AIC, smoothness priors and linear-Gaussian smoothness priors state space modeling and some of its applications are briefly reviewed. In the applications we present trend estimation, seasonal time series estimation, and the estimation of a time-varying autoregressive model for the estimation of gradually changing spectra. The general state space model is treated in Section III. Included are brief treatments of the numerical approximation, Gaussian sum approximation and Monte Carlo methods of general state space filter/smoother implementations. Finally, the emphasis of the chapter, on illustrating applications of general state space modeling is in Section IV. In the applications of mostly real data analysis, we present abrupt trend estimation, abrupt seasonal estimation, and estimation of abruptly changing variance and abruptly changing spectrum estimation. Also included are the modeling of an inhomogeneous discrete process, of a quasi-periodic process, the filtering, and the smoothing of a nonlinear process. The examples are realized by the different methods of general state space implementation. For the entire collection see [Zbl 0894.00023].