About stochastic integrals with respect to processes which are not semi- martingales (Q762832)

We consider on the one hand a family \(Z^{\alpha}\) of semi-martingales on (\(\Omega\),\({\mathcal F}_ t,{\mathbb{P}})\) depending on a real parameter \(\alpha\), and on the other hand a process U with orthogonal increments defined on an other probability space (W,\({\mathcal W},{\mathbb{Q}})\). If we mix the semi-martingales \(Z^{\alpha}\) by integration in \(\alpha\) with respect to U, we get a process \[ Z_ t(\omega,w)=\int_{\alpha \in {\mathbb{R}}}Z_ t^{\alpha}(\omega)dU_{\alpha}(w) \] such that the stochastic integral \(\int^{t}_{0}H_ s(\omega)dZ_ s(\omega,w)\) exists for any bounded predictable H on (\(\Omega\),\({\mathcal F}_ t,{\mathbb{P}})\) and defines a \(\sigma\)-additive vector measure with values in \(L^ 2({\mathbb{P}}\times {\mathbb{Q}})\) although it may be that for \({\mathbb{Q}}\)- almost all w the process \(Z_ t(.,w)\) is not a semi-martingale on (\(\Omega\),\({\mathcal F}_ t,{\mathbb{P}})\).