Finite sample properties of system identification of ARX models under mixing conditions (Q5926264)

An input-output system depending on unknown parameters is considered. In particular, the author considers ARX models with a uniformly bounded (identification) criterion function. In the literature, great attention was paid to the identification of the above-mentioned systems. Especially the asymptotic properties were investigated under the assumptions of independent random data. The aim of the author is to focus on the dependent case and, moreover, to obtain uniform probabilistic bounds based on the difference between the expected value of the identification criterion and the empirical value evaluated on the observed data. In particular, the author investigates the properties of the prediction error for a relatively small size of random samples in the case of \(\beta\)-mixing. A possible generalization to the \(\Phi\)-mixing case is mentioned. Evidently, the presented assertions generalize the known results for the independent case. The paper is completed by valuable references on mixing processes.