Some new results in the theory of recursive identification

A description is given of a few basic concepts such as L-mixing processes, and a few improved asymptotic results on offline estimators are presented. This is followed by a rate-of-convergence result for L. Ljung´s scheme (1976, 1977) and strong approximation theorem which couples offline and online estimators. These previous results are applied to get the best result to date in the asymptotic theory for J. Rissansen´s predictive stochastic complexity (1986, 1987, 1989). A brief review is presented of what can be done with slowly time-varying systems