No label defined (Q4940295)

It is proved that the convergence of random vectors in the Sobolev space and non-degeneracy of the Malliavin matrix implies the convergence in variation of distributions. Let \(f_{n}, f:B\to R^{d}\), where \(B=\{ x\in R^{d} \mid \|x\|\leq 1\}.\) Let \(W_{p}^1(B,R^{d})\) be the Sobolev space which is the completion of \(C^1(B,R^{d})\) by the norm \[ \|g\|_{W_{p}^1}= \|g\|_{L_{p}(B,\lambda^{d},R^{d})}+ \sum\limits_{i=1}^{d} \|\nabla g^{i}\|_{L_{p}(B,\lambda^{d},R^{d})}, \quad g=(g^1,\ldots,g^{d}). \] The following assertion is the main result of this paper: Let \(f_{n}\to f\) as \(n\to\infty\) in \(W_{d}^1(B,R^{d})\), \(L\subset \{ x\in B\mid\det \nabla f(x)\not=0\}\), \(L\) is a Borel set. Then \( \lambda_{L}^{d}\circ f_{n}^{-1}@>\text{Var}>>\lambda_{L}^{d}\circ f^{-1}\) as \(n\to\infty\). Here \(\lambda_{L}^{d}\) is a Lebesgue measure on the set \(L.\) As examples for the integral functionals the author considers the multiple Wiener integrals and finite-dimensional distributions of diffusion processes. For example, the following statement is proved: Let \(\{\Omega,{\mathcal F},P\}\) be a probability space, let \(\xi\) be a random element in a Polish space \(X\), and let \(\sigma(\xi)\) be a \(\sigma\)-algebra generated by \(\xi\). Let \(F_{n}, F:\Omega\to X\) be random elements. Suppose that for all measurable sets \(L\subset X\), \(P_{L}^{F_{n}}:=P_{L}\circ F_{n}^{-1}\), the distribution of \(F_{n}\) in \(X\), converges in variation to \(P_{L}^{F}\) and \(F_{N}\to F\) as \(n\to\infty\) in probability. Then for every random variable \(\xi\in L_{p}(\Omega,{\mathcal F},P)\), \(p\geq 1\), the convergence of conditional mathematical expectations \(E\{\xi\mid \sigma(F_{n})\} @>L_{p}>> E\{\xi\mid \sigma(F)\}\), \(n\to\infty\), holds.