scientific article; zbMATH DE number 1064827
For iid random variables \(X_1, \dots, X_n\) and a map \(\Phi\) such that \[ \Phi (x_1,x_2)= \Phi (x_2,x_1),\;E\Phi (X_1,X_2)=0\;\text{and} E\|\Phi (X_1, X_2) \|< \infty \] with norm \(\|\cdot\|\), we define the \(U\)-statistic with kernel \(\Phi, U_n= {n\choose 2}^{-1}\sum_{1<j\leq n}\Phi (X_i, X_j)\), and represent the Hoeffding decomposition as \[ U=2n^{-1} \sum^n_{j =1} g_j(X_j) +2n^{-1} (n-1)^{-1} \sum_{1<j \leq n}g_2 (X_i,X_j), \] where \(g_1(x)= E\Phi (x,X_1)\), \(g_2(x,y) =\Phi(x,y) -g_1(x) -g_2(y)\). Finally, let \(Y\) be a Gaussian \(H\)-valued random variable with mean zero and covariance operator \(\sigma^{-2}V\), where \(0<\sigma^2 =E\|g_1(X_1) \|^2 <\infty\), and \(V\) denotes the covariance operator of \(g_1(X_1)\). The author derives the Berry-Esseen type estimate \[ \begin{multlined} \Delta_n(a)= \sup_{r\geq 0} \biggl|P\bigl (\|2^{-1} n^{1/2} \sigma^{-1} U_n-a \|\leq r\bigr) -P\bigl(\|Y-a \|\leq r\bigr) \biggr|\leq \\ \leq c\bigl(1 +\|a\|^3 \bigr)\sigma^4 \left( \prod^9_{j=1} \sigma_j^{-1} \right) \biggl(E\bigl\|\Phi(X_1,X_2) \bigr \|^3 \biggr)^2n^{-1/2}.\end{multlined} \] He also considers the Berry Esseen type estimate on the bootstrapping of Von Mises \(w^2\)-statistics, where \[ w^2_n= n\int^\infty_{- \infty} \bigl(F_n(x) -F(x)\bigr)^2 dF(x), \quad F(x)= P(X_1\leq x),\quad \text{and} \quad F_n(x)= n^{-1} \sum^n_{j=1} 1\{X_j\leq x\}. \] For the entire collection see [Zbl 0856.00029].
- scientific article; zbMATH DE number 3942610 (Why is no real title available?)
- scientific article; zbMATH DE number 4100390 (Why is no real title available?)
- Normal Approximation of U-Statistics in Hilbert Space
- Bootstrap for dependent Hilbert space-valued random variables with application to von Mises statistics
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 Q4356176)