Expansions for U-statistics and von Mises functionals (Q1083782)

The authors establish a one-term Edgeworth expansion for a non-degenerate U-statistic of arbitrary but fixed degree m based on n i.i.d. r.v.'s \(X_ 1,...,X_ n\), with a remainder term of order \(n^{-3/5}\). The assumptions needed for this are a third absolute moment for the kernel h and a fourth moment for the conditional expectation g of the kernel, given one of its arguments. In addition, Cramer's condition (C) is imposed on \(g(X_ 1).\) Secondly, the problem of establishing expansions for the moments of a normalized sum of independent and identically distributed random variables, taking values in a separable real Hilbert space is investigated, using von Bahr's method [\textit{B. von Bahr}, Ann. Math. Stat. 36, 808-818 (1965; Zbl 0139.353)]. This latter method employs a very useful integral representation for the moments under consideration, which involves the characteristic function (ch.f.) of the normalized sum of i.i.d. r.v.'s. The main difficulty is then to find suitable estimates for the ch.f. for large values of its argument. The relation between even moments of normalized sums and von Mises functionals with an appropriate kernel is also pointed out.