Vector valued stochastic processes. IV: Integral representation of linear operations on spaces of stochastic processes (Q910814)

[For part III see Stochastic processes, Semin., Princeton/New Jersey 1987, Prog. Probab. Stat. 15, 93-122 (1988; Zbl 0647.60048)]. Let (\(\Omega\),F,P) be a probability measure space, \((F_ t)_{t\geq 0}^ a \)filtration satisfying the usual conditions, E,F Banach spaces and \(1\leq p<\infty\). We denote by \(R^ p_ E\) the vector space of all E-valued, cadlag, adapted processes X with separable range, having limit at \(\infty\) and satisfying \(\| X\|_{R^ p}:=\| X^*\|_{L^ p}<\infty\) where \(X^*=\sup_{t\geq 0}| X_ t|\). With \(\| X\|_{R^ p}\) as a norm, \(R^ p_ E\) is a Banach space. For a linear continuous mapping J: \(R^ p_ E\to F\) we denote by \(\| J\|\) the usual norm and we define \(\|| J\|| =\sup\) \(\sum_{i\in I}| J(X^ i)|\), where the supremum is taken for all finite families \((X^ i)_{i\in I}\) from \(R^ p_ E\) with \(\| \sum_{i\in I}\| X^ i\| \|_{R^ p}\leq 1.\) If \(F={\mathbb{R}}\) then \(\|| J\|| =\| J\|\). Define also \(| J|:\) \(R^ P_{{\mathbb{R}}}\to {\mathbb{R}}\) by \(| J| (| \phi |)=\sup\) \(\sum_{i\in I}| J(X^ i)|\), where \(X^ i\) are as above, with \(\sum_{i\in I}| X^ i| \leq | \phi |\). Then \(\| | J| \| =\|| J\||.\) Theorem: If \(\|| J\|| <\infty\), then we have \[ J(X)=\int X_- dm_-+\int Xd m_+\quad for\quad X\in R^ p_ E\quad and\quad | J| (\phi)=\int \phi_-d| m_-| +\int \phi d| m_+| \quad for\quad \phi \in R^ p_{{\mathbb{R}}}, \] where \(m_-\) (respectively \(m_+)\) is an L(E,F)-valued predictable (respectively optional) measure with finite variation. If E has the Radon-Nikodym property, we have also \[ J(X)=E(\int X_{s-}dA^-_ s+\int X_ sdA^+_ s)\quad and\quad | J| (\phi)=E(\int \phi_{s-}d| A^-|_ s+\int \phi_ sd| A^+|_ s), \] where \(A^-\) (respectively \(A^+)\) is an \(L(E,F'')\) valued, predictable (respectively optional) process with integrable variation.