Asymptotic Analysis for U-Statistics and Its Application to Von Mises Statistics英文文献资料.docVIP

Open Journal of Statistics, 2011, 1, 139-144 139 doi:10.4236/ojs.2011.13016 Published Online October 2011 (http://www.SciRP.org/journal/ojs) Asymptotic Analysis for U-Statistics and Its Application to Von Mises Statistics Timur Zubayraev Faculty of Computational Mathematics and Cybernetics, Moscow State University, Moscow, Russia E-mail: tzubayraev@ Received July 30, 2011; revised August 31, 2011; accepted September 14, 2011 Abstract Let X , X , X1,?, X N be i.i.d. random variables taking values in a measurable space (X，B). Let ?1: X ?? and ? : X ?? be measurable functions. Assume that ? is symmetric, i.e. ??x, y?=?? y, x?, for 2 any x, y?X . Consider U-statistic T = N1 1 ? ?(Xi, X j ) ? 1?i j?N ? ?1(Xi) , assuming that ??1(X ) = 0 , 1?i?N N ??(x, X ) = 0 for all x?X , ?? (x, X ) ? , ??1 (X ) ? . We will provide bounds for ?N = supx F(x) 2 2 ?F0(x) ? F1(x) , where F is a distribution function of T and F0 , F are its limiting distribution func- 1 tion and Edgeworth correction respectively. Applications of these results are also provided for von Mises statistics case. Keywords: U-Statistics, Von Mises Statistics, Symmetric Statistics 1. Introduction ?1) . Thus, we have ?1(x) = ?a je j (x) in L2 , ?2 =??12(X ) = ? a j?0 2 j , (1.2) Consider the measurable space (X，B, ? ), with meas- ure ? =?(X) . Let L = L (X，B, ?) denote the real 2 j?0 with a j =??1(X )ej (X ) and ?ej (X ) = 0 , for all j. Therefore (e j (X )) j?0 is an orthonormal system of ran- dom variables with zero means. Hilbert space of square integrable real functions. Let ? : L ? L denote the Hilbert-Schmidt operator asso- 2 2 ciated with the kernel ? and defined via Hilbert space ?? , such that ?2 ?? consists of x = (x1, x2?)? ? ?f (x) = ?X?(x, y) f (y)?(dy) =??(x, X ) f (X ) , ? Let q ,q ? be its eigenvalu