Mixing properties of ARMA processes (Q1104632)

Let \(\{\) Y(t)\(\}\) be a stationary process in R l. Denote by \({\mathcal A}_ 0\) and \({\mathcal A}_ k\) the \(\sigma\)-algebra generated by \(\{\) Y(t),t\(\leq 0\}\) and \(\{\) Y(t),t\(\geq k\}\), respectively. Define \[ \beta (k)=E\sup_{B\in {\mathcal A}\quad k}| P(B| {\mathcal A}_ 0)- P(B)|. \] If there exists \(\rho\in (0,1)\) such that \(\beta (k)=O(\rho\) k), the process \(\{\) Y(t)\(\}\) is called geometrically completely regular (GCR). The author proves that stationary vector ARMA processes are GCR (and hence strong mixing) if the innovations have absolutely continuous distribution with respect to Lebesgue measure.