\(U\)-max-statistics (Q953861)

A \(U\)-statistic with \(k\)-variate symmetric kernel \(h\) is the average of the \(n \choose k\) values obtained by evaluating \(h\) at all possible subsets of size \(k\) drawn from the random sample \(\xi_1, \ldots, \xi_n .\) The \(U\)-max statistic, \(H_n,\) is defined to be the maximum of the \(n \choose k\) kernel values \(h(\xi_{i_1}, \ldots, \xi_{i_k}), \;1 \leq i_1 < \ldots < i_k \leq n.\) This paper applies Theorem 2.N of \textit{A. Barbour, L. Holst} and \textit{S. Janson} in `Poisson Approximation.' Oxford: Clarendon Press (1992; Zbl 0746.60002), to obtain a limit law for \(H_n\). This result is then applied to obtain the asymptotic distributions and rates of convergence for the largest interpoint distance within a sample of points in \(R^d\) and the largest scalar product. A further application of the main result is to the largest perimeter of all triangles formed by triplets of independent and uniformly distributed points on the unit circle.