Asymptotics for linear processes (Q1193361)

A new method of deriving asymptotics for linear processes of the form \(X_ t=\sum c_ j\varepsilon_{t-j}\), where \(\{\varepsilon_ t\}\) is a sequence of time series innovations, is developed. The method is based on an algebraic decomposition of the linear filter into long-run and transitory components which is due to \textit{Beveridge} and \textit{Nelson} [J. Monetary Economics 7, 151-174 (1981)]. This method reduces a given limit problem for a linear process to the same problem for innovations \(\varepsilon_ t\) and to an estimation of some remainder. The authors present in a unified way a wide spectrum of limit results like laws of large numbers, laws of iterated logarithm, central limit theorems and invariance principles. The results accomodate both homogeneous and heterogeneous innovations as well as innovations with undefined means and variances. Besides results known from the literature some new theorems are also proved.