Time‐varying parameters and state space models
Timo Teräsvirta, Dag Tjøstheim, W. J. Granger · 2010
Abstract Linear state space models have become popular in time series, and there are applications to many fields. The Kalman filter is often a fundamental tool. In this chapter it is shown that there are extensions of these concepts to a nonlinear framework through such devices as the extended Kalman filter and particle filters. Hidden Markov chains represents an alternative but related technique, where parameters are replaced by stochastic processes; i.e., Markov chains. The chapter also contains a short section on estimating these types of models.