Estimating the system order by subspace methods

This paper discusses how to specify the order of a state-space model. To do so, we start by revising existing approaches and find in them two basic shortcomings: (i) some of them have a poor performance in short samples and (ii) most of them are not robust, meaning that their performance critically depends on the data generating process. We tackle these two issues by proposing new and refined criteria. Monte Carlo simulations provide evidence of the potential of the proposals. Copyright Springer-Verlag 2012