Sequential Bayesian Inference for Dynamic State Space Model Parameters

A method for sequential Bayesian inference of the static parameters of a dynamic state space model is proposed. The method uses filtering and prediction distribution approximations such as the extended and unscented Kalman filters; any other valid approximation can also be used. The method computes the posterior distribution on a discrete grid that tracks the support dynamically. Simulation studies show that the method provides a good trade off between computation speed and accuracy, relative to the integrated nested Laplace approximation and a particle filter, in examples of both non-linear and non-Gaussian models.