Abstract:
Let $X_1,\dots,X_n$ be independent identically distributed random variables. An optimal Lyapunov (or Berry–Esseen) bound is derived for $U$-statistics of degree 2, that is, statistics of the form $\sum_{j<k}H(X_j,X_k)$, where $H$ is a measurable, symmetric function such that $\mathbf{E}\,|H(X_1,X_2)|<\infty$, assuming that the statistic is nondegenerate.
Keywords:$U$-statistics, Lyapunov-type bound, Berry–Esseen bound, rate of convergence, normal approximations.