engleski

On one-step extrapolation error of vector time series

The asymptotic of the mean-square prediction error is studied for vector weakly stationary time series when the spectral densities of coordinate processes vanish on an interval of positive Lebesgue measure in the so-called on-step extrapolation procedure. We show that the convergence rate of the prediction error approaches zero exponentially, similarly as in the scalar case.

Finite predictor ; Singular stochastic process ; Vector stochastic process ; Cross-spectral density ; One-step extrapolation

nije evidentirano