A Note on Optimal Smoothing for Time Varying Coefficient Problems
NBER Working Paper No. 128
An algorithm is presented which provides a complete solution to the optimal estimation problem for time-varying parameters when no proper prior distribution is specified. The key ideas involve a combination of the information-form Kalman filter with the two-filter interpretation of the optimal smoother. The algorithm produces efficient estimates of the parameter trajectories over the entire sample, arid is equally applicable when a proper prior distribution has been specified.
Document Object Identifier (DOI): 10.3386/w0128