scientific article; zbMATH DE number 4111802

From MaRDI portal
Publication:3834870





Given \(X_ 1,X_ 2,...,X_ n\) i.i.d. r.v. and \(G(x_ 1,...,x_ m)\) a symmetric m-dimensional kernel function (m\(\leq n)\) the authors consider the U-statistics \[ U_ n=\left( \begin{matrix} n\\ m\end{matrix} \right)^{- 1}\sum_{1\leq i_ 1<...<i_ m\leq n}G(X_{i_ 1},...,X_{i_ m}),\quad n\geq 2. \] The fact, that the set of all non-degenerate limit df of normalized \(U_ n\) coincides with a subset of all stable df \(G_{\alpha}\), \(\alpha\in (1,2]\), was obtained by \textit{T. L. Maleviĉ} and \textit{B. Abdalimov} [Theory Probab. Appl. 22, 370-377 (1978); translation from Teor. Veroyatn. Primen. 22, 379-386 (1977; Zbl 0387.60028)]. Here the authors investigate the rate of convergence of the df of normalized U-statistics to a given stable law \(G_{\alpha}\), \(\alpha\in (1,2]\), under some pseudomoment conditions. In case \(\alpha =2\), the obtained rate will be of order \(n^{-1/3}\) what is not better than the known one. However in case \(1<\alpha <2\) their results seem to be new.











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 Q3834870)