Strong asymptotic independence on Wiener chaos (Q2789883)

In this paper, the asymptotic independence of the components of random vectors \((F_{1,n}, \ldots, F_{d,n})\) is studied, where every sequence \((F_{j,n})_{n \in \mathbb N}\) is contained in a fixed Wiener chaos. In the main result it is proved that for variance one elements, \(q\geq 1\), the highest order of Wiener chaos, smooth test functions \(\psi_j \in C^{\infty}_q(\mathbb R)\) (i.e., infinitely differentiable, all partial derivatives up to the order \(q\) are bounded) and some constant \(c\), the following inequality holds: NEWLINE\[NEWLINE \left|\mathbb{E}\left[\prod_{j \leq d} \psi_j(F_{j,n})\right] - \prod_{j \leq d} \mathbb{E}\left[\psi_j(F_{j,n})\right]\right| \leq c\left(\|\psi_d'\|_{\infty} \prod_{j \leq d-1} \|\psi_j\|_{q}\right)\sum_{1 \leq i <j \leq d} \text{Cov}(F_{i,n}^2,F_{j,n}^2). NEWLINE\]NEWLINE This yields that the asymptotic independence of the sequence is equivalent to NEWLINE\[NEWLINE \lim_{n \rightarrow \infty} \text{Cov}(F_{i,n}^2, F_{j,n}^2) =0 NEWLINE\]NEWLINE for all \(i \neq j\). The proof makes use of Malliavin calculus, in particular of the Meyer inequalities.