Total variation distance between stochastic polynomials and invariance principles (Q2189458): Difference between revisions

Suppose that a sequence \(X=(X_n)_{n \in \mathbb N}\) consists of independent random variables, which have finite moments of any order. This paper considers stochastic polynomials \[Q_{N,k_ *}(c,X) = \sum\limits_{m = 0}^N\Phi _m(c,X),\] where \[\Phi_m(c,X) := \sum\limits_{k_1,\dots,k_m=1}^{k_ *}\sum\limits_{n_1<\dots <n_m=1}^\infty c((n_1,k_1),\dots,(n_m,k_m)) \times \mathop \Pi \limits_{j = 1}^m (X_{n_j}^{k_j}-\mathrm{E}(X_{n_j}^{k_j})).\] The coefficients \(c\) are symmetric and null on the diagonals and only a finite number of them are nonnull. Here \(X_n\in\mathbb{R}\), but the paper deals with\(X_n\in\mathbb{R}^{d_ *}\). The authors indicate that these multilinear stochastic polynomials are a natural generalization of elements of the classical Wiener chaos. In addition, they are of interest in applications to \(U\)-statistics theory. The goal of the paper is to estimate the total variation distance between the laws of two such polynomials, and to establish an invariance principle.