1: \begin{abstract}

2: It is commonly acknowledged that V-functionals with an unbounded kernel

3: are not Hadamard differentiable and that therefore the asymptotic

4: distribution of U- and V-statistics with an unbounded kernel cannot be

5: derived by the Functional Delta Method (FDM). However, in this article

6: we show that V-functionals are quasi-Hadamard differentiable and that

7: therefore a modified version of the FDM (introduced recently in

8: (\textit{J.~Multivariate Anal.} \textbf{101} (2010) 2452--2463)) can

9: be applied to this problem. The modified FDM

10: requires weak convergence of a weighted version of the underlying

11: empirical process. The latter is not problematic since there exist

12: several results on weighted empirical processes in the literature; see,

13: for example, (\textit{J.~Econometrics}

14: \textbf{130} (2006)

15: 307--335,

16: \textit{Ann. Probab.}

17: \textbf{24} (1996)

18: 2098--2127,

19: \textit{Empirical Processes with Applications to Statistics} (1986)

20: Wiley,

21: \textit{Statist. Sinica}

22: \textbf{18} (2008)

23: 313--333). %\cite{ChenFan2006,ShaoYu1996,ShorackWellner1986,Wu2008}

24: The modified FDM approach has the advantage that it is very flexible

25: w.r.t. both the underlying data and the estimator of the unknown

26: distribution function. Both will be demonstrated by various examples.

27: In particular, we will show that our FDM approach covers mainly all the

28: results known in literature for the asymptotic distribution of U- and

29: V-statistics based on dependent data -- and our assumptions are by

30: tendency even weaker. Moreover, using our FDM approach we extend these

31: results to dependence concepts that are not covered by the existing literature.

32: \end{abstract}