scientific article; zbMATH DE number 4013634
For a probability distribution F and some measure G, which are defined on a class \({\mathcal H}\) of Borel sets of a separable Hilbert space H denote \(\tilde F(A)=F(A\cap O_ 1)+G(A\cap O^ c_ 1)\), \(A\in {\mathcal H}\), where \(O_ R\) is a closed sphere of radius R with a center at the zero point \(\theta\) of the space H. One can say that F is ``corrected by the measure G and call \(\tilde F\) the ``corrected distribution (when \(G(O^ c_ 1)=0\), \(\tilde F\) is a truncation of the distribution \(\tilde F\)). Let \(F_ k\), \(k=1,...,n\), be probability measures on \({\mathcal H}\) with zero expectations and covariance operators \(B_ k\) such that \(sp(B_ k)\neq 0\) and \(sp(B_ 1+...+B_ n)=1\). Denote \(F=*^{n}_{k=1}F_ k\), \(\bar F=*^{n}_{k=1}\tilde F_ k\), where \(\tilde F_ k\) is a distribution ``corrected by some measure \(G_ k\), \(V_ k\) is a full variation of the generalized measure \(F_ k-G_ k\), \(V_ k(R)=V_ k(O^ c_ R)\) and \[ C_ G(m) = \max \{sp(B_ k)^{-\ell /2}\int_{| x| \geq 1}| x|^{\ell}G_ k(dx): k\leq n,\quad 0\leq \ell \leq m\}. \] For any Borel \(A\not\ni \theta\) and any \(m\geq 2\) it is proved that \[ | (F-\bar F)(A)| \leq C_ G(m)\{\sum^{n}_{k=1}V_ k(\max (1,\beta /2))+ \] \[ (1+\beta)^{- m}[\sum^{n}_{k=1}V_ k(1)][1+\sum^{n}_{k=1}\int_{1\leq | x| \leq 1+\beta}| x|^ m V_ J(dx)]\}, \] where \(\beta\) is the minimal distance from \(\theta\) to the boundary \[ \bar C_ G(m)=(C_ G(m)+C(m))(1+C_ G(2))^ m\exp [C_ G(2)] \] and the constant \(C(m)\) is but m-dependent.
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 Q3025951)