Gaussian and bootstrap approximations for high-dimensional U-statistics and their applications (Q1750282)

Given independent and identically distributed random vectors \(X_i\), \(i=1,\dots,n\), and a fixed measurable function \(h=h(x_1,x_2)\), with \(h(x_1,x_2) = h(x_2,x_1)\), consider the uniform approximation of the \(U\)-statistics \(U_n\) of order two over the class \(\mathcal{A}\) of all hyperrectangles \(A\) in \(\mathbb{R}^d\). Hence, for \(\theta := Eh(X_i,X_j)\) and \(T_n = \sqrt{n} \frac{U_n - \theta}{2}\), the problem is the uniform approximation of the probability \(P(T_n \in A)\) by \(P(\widetilde{T}_n \in A)\) with a certain random vector \(\widetilde{T}_n\) in \(\mathbb{R}^d\) as well as to determine corresponding error bounds. Moreover, the dependence of the asymptotic behavior of the \(U\)-statistics on the dimension \(d\) is studied.