U-statistics in characterization problems (Q808568)

Let \(x_ 1,...,x_ n\) be independent random variables (r.v.s.) with the same distribution function (d.f.) F(x). The static \[ U(x_ 1,...,x_ n)=\sum u(x_{i_ 1},...,x_{i_ k}), \] where the summation is over the set \(I_ k\) of values of the multiindex \((i_ 1,...,i_ k)\) satisfying the condition \(1\leq i_ 1\leq n,...,1\leq i_ k\leq n\), is, by definition, a U-statistic of rank k with the kernel \(u(z_ 1,...,z_ k)\). Since the publication of \textit{W. Hoeffding}'s classical paper [Ann. Math. Stat. 19, 293-325 (1948; Zbl 0032.04101)], U-statistics have become a popular object of research in probability theory and mathematical statistics, mainly in the asymptotic setting as \(n\to \infty.\) The aim of this note is to show that U-statistics are also natural and useful in typical characterization problems of mathematical statistics for a fixed n.