Square Root Iterative Filter: Theory and Applications to Econometric Models

This paper provides a new algorithm for estimating state space dynamic models and, as an example, it considers the estimation of time-varying parameter models. The novel elements of the algorithm are: a simple, easily implementable, square root method which is shown to solve the numerical problems affecting the standard Kalman filter algorithm and the related information filter and smoothing algorithms;an iterative framework, where information and covariance filters and smoothing are sequentially run in order to estimate all the parameters of the model; four different algorithms to consistently estimate the distribution of the estimated parameters, which are described and then compared by performing appropriate Montecarlo experiments.