Filtering and smoothing algorithms for state space models

AbstractThe paper reviews and generalizes recent filtering and smoothing algorithms for observations generated by a state model. In particular the paper discusses the modified Kalman filter derived by Ansley and Kohn (1985) and Kohn and Ansley (1986) to deal with state space models having partially diffuse initial conditions, and shows how to compute the limiting normalized likelihood of the observations for such models. The paper also discusses and generalizes the new smoothing algorithm presented by Kohn and Ansley (1987c, 1989) and extends it to state space models with partially diffuse initial conditions