scientific article; zbMATH DE number 3885165

From MaRDI portal
Publication:3218972





[For the entire collection see Zbl 0542.00014.] The paper considers stationary continuous-time processes with rational spectral densities, which are observed at discrete time instants. The problem treated is concerned with fitting continuous-time rational models to discrete time data. The basic idea is to introduce a (discrete-time) ARMA model which is equivalent to the continuous-time rational model in the sense that the correlation functions of the two models take the same values at the sampling points. The ARMA model is parameterized by the AR parameters and the few first covariances of the series of data. Alternatively, it may be parameterized by the AR parameters and the MA covariances. Since the parameters of the ARMA model are functions of the parameters of the equivalent continuous-time model, the discrete data were parameterized in terms of the latter set of parameters. As shown in the paper this parameterization turns out to be well suited to: a) Discussion of the aliasing problem. b) Approximating and exact likelihood construction. c) Generation of interpolates, derivatives and forecasts.











This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3218972)