Orthonormal Banach systems with applications to linear processes

From MaRDI portal





A sequence of random variables \((X_ n,n\in {\mathbb{Z}})\) is called a convergence system if, for every sequence of constants \(a_ n\), \(\sum a^ 2_ n<\infty \Rightarrow \sum a_ nX_ n\) converges a.e.. It is called a strict convergence system if \(\Rightarrow\) is replaced by \(\Leftrightarrow\). It is called a Banach system if there exists \(C>0\) such that \(E| \sum^{n}_{i=m}a_ iX_ i| \geq C(\sum^{n}_{i=m}a^ 2_ i)^{1/2}\) holds for all \(n\geq m\) and all constants \((a_ i).\) The authors consider martingale difference sequences \((X_ n)\) with \(EX^ 2_ n=1\) and inf \(E| X_ n| >0\). Let \((b_ n)\) be a sequence of constants with \(\sum b^ 2_ n<\infty\), and \(Y_ n=\sum^{\infty}_{i=-\infty}b_{n-i}X_ i\). Let f denote the spectral density of the wide sense stationary process \((Y_ n)\). It is shown that if ess sup f\(<\infty\) and ess inf f\(>0\), then \((Y_ n)\) is a Banach system and, for any sequence \((a_ n)\), the convergence in distribution of \(\sum a_ nX_ n\) implies \(\sum a^ 2_ i<\infty\). Moreover, if sup \(E| X_ n|^ p<\infty\) for some \(p>2\), then (i) \((Y_ n)\) is a convergence system iff ess sup f\(<\infty\), and (ii) \((Y_ n)\) is a strict convergence system iff ess inf f\(>0\) and ess sup f\(<\infty\), and also iff \((Y_ n)\) is a convergence system and a Banach system. Some of the results hold also for linear processes \((Y_ n)\) generated by orthogonal Banach systems, and \(S_ p\)-systems.











This page was built for publication: Orthonormal Banach systems with applications to linear processes

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3692557)