Whittle estimation for stationary state space models with finite second moments

Abstract

In this paper, we consider the Whittle estimator for the parameters of a stationary solution of a continuous-time linear state space model sampled at low frequencies. In our context the driving process is a L\'evy process which allows flexible margins of the underlying model. The L\'evy process is supposed to have finite second moments. It is well known that then the class of stationary solutions of linear state space models and the class of multivariate CARMA processes coincides. We prove that the Whittle estimator, which is based on the periodogram, is strongly consistent and asymptotically normally distributed. A comparison with the classical setting of discrete-time ARMA models shows that in the continuous-time setting the limit covariance matrix of the Whittle estimator has an additional correction term for non-Gaussian models. For the proof, we investigate as well the asymptotic normality of the integrated periodogram which is interesting for its own. It can be used to construct goodness of fit tests. Furthermore, for univariate state space processes, which are CARMA processes, we introduce an adjusted version of the Whittle estimator and derive as well the asymptotic properties of this estimator. The practical applicability of our estimators is demonstrated through a simulation study.