scientific article; zbMATH DE number 4036816

From MaRDI portal
Publication:3776310





The paper deals with decompositions of strictly stationary sequences of random variables. The process is represented by \((f\circ T^ i)_{i\in {\mathbb{Z}}}\) where f is a measurable function on a probability space (\(\Omega\),\({\mathcal A},\mu)\) and T is a measure-preserving mapping of \(\Omega\) onto itself. A \(\sigma\)-algebra \({\mathcal M}\subset {\mathcal A}\) is called invariant if \(T^{-1}{\mathcal M}\subset {\mathcal M}\); the maximal invariant \(\sigma\)-algebra \({\mathcal P}\) such that \(T^{-1}{\mathcal N}={\mathcal N} mod \mu\) for each invariant \(\sigma\)-algebra \({\mathcal N}\subset {\mathcal P}\) is called the Pinsker \(\sigma\)-algebra. For invariant \(\sigma\)- algebras \({\mathcal M}\subset {\mathcal A}\) and \({\mathcal Q}\subset {\mathcal P}\) and an integrable \({\mathcal M}\)-measurable function f it holds \[ E(f| T^{- 1}{\mathcal M})=E(f| {\mathcal Q}\vee T^{-1}{\mathcal M})\quad a.s. \] Using this equality it is proved that the Hilbert space \(L^ 2({\mathcal A})\Theta L^ 2({\mathcal P})\) is the space of the square integrable functions f for which \((f\circ T^ i)_{i\in {\mathbb{Z}}}\) is a sum of at most countably many martingale difference sequences. The decomposition is preserved for ergodic components of \(\mu\), and if \((f\circ T^ i)_{i\in {\mathbb{Z}}}\) is a sequence of martingale differences w.r. to \(\mu\) then it is a sequence of martingale differences w.r. to almost all ergodic components of \(\mu\). The results can be applied to martingale limit theorems for strictly stationary processes.












This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3776310)