Estimating time varying state correlations in state space models using indirect inference.
In this paper, a multivariate state space model is proposed where the correlations between the innovations of the state variables are assumed to be time dependent. The model is fitted via indirect inference where cubic splines are used as an auxiliary model and a bootstrap filter is used for estimating the time-dependent correlations and the other state variables. The method is applied to a state space model where monthly unemployement figures observed via the Dutch Labour Force Survey are combined with series of claimant counts.
