Maximum standardized cumulant deconvolution of non-Gaussian linear processes (Q751716)

The author considers a linear process \(x_ t=\sum_{s\in G}u_ sw_{t-s}\), where \(u=\{u_ t\}\) is a strict white noise, G is a countable Abelian group and the system response sequence \(w=\{w_ t\}\) satisfies \(0<\sum_{t\in G}| w_ t|^ 2<\infty\). The concept of natural peakedness of w is defined and investigated. The convergence theory of maximum standardized cumulant deconvolution is presented and a relation between cumulant deconvolution and autoregressive models on G is given.