scientific article; zbMATH DE number 3942610

From MaRDI portal
Publication:3713249





Let \(\{X_ n\}\) be a sequence of i.i.d. random variables and \(\phi\) (x,y) a bivariate symmetric function with E \(\phi\) (X\({}_ 1,X_ 2)=0\). Define U-statistics \[ (1)\quad U_ n=\left( \begin{matrix} n\\ 2\end{matrix} \right)^{-1}\sum_{1\leq i<j\leq n}\phi (X_ i,X_ j). \] We extend the results by \textit{H. Callaert} and \textit{P. Janssen} [Ann. Stat. 6, 417- 421 (1978; Zbl 0393.60022)] and by \textit{L. Zhao} [see Chin. Ann. Math., Ser. A 4, 699-706 (1983; Zbl 0509.60024)] to more general cases. Let g(x) be any function satisfying: (a) it is nonnegative, even and non- decreasing in the interval \(x>0\); (b) x/g(x) is non-decreasing in the interval \(x>0\), and \(x^ 2g(x)\) is convex. Suppose that \(\psi (X_ 1)=E[\phi (X_ 1,X_ 2)| X_ 1]\) has a positive variance and \[ E[\phi^ 2(X_ 1,X_ 2)g(| \phi (X_ 1,X_ 2)|)]<\infty. \] Then there exists a constant c such that \[ \sup_{x}| P(\sigma_ n^{-1}U_ n<x)-\Phi (x)| \leq cg^{-1}(\sqrt{n}), \] where \(\sigma^ 2_ n=EU^ 2_ n\) (Theorem 1). We may further generalize this result to the case that only the existence of second moment is required (Theorem 2). For von Mises statistics, we obtain a conclusion (Theorem 3) corresponding to theorem 1.











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 Q3713249)