A covariance extension approach to identification of time series (Q1975568)

In this beautiful paper, the authors analyze a three-step procedure for identification of time series. First, a partial covariance sequence based on a finite time series is constructed. Second, using the maximum entropy extension, an AR-model with degree \(m\) is constructed. Third, via a stochastic model reduction procedure, an \(n\)-degree approximation, with \(n\) much smaller than \(m\), is determined. Now follows the real meat of the paper. Supposing the data is generated from a true finite-dimensional system that is minimum phase, it is shown that the transfer function of the estimated system tends in \({\mathcal H}^\infty\) to the true transfer function as the length of the time series tends to infinity. The proposed procedure, and some of its variations, are evaluated by simulations.