Normal approximation for finite-population U-statistics

Let \(X_ 1,...,X_ n\) be a random sample without replacement from the set of real numbers \(\{a_{N1},...,a_{NN}\}\). Define the U-statistic \(U=U(N,n)\), for fixed m, by \(U=\left( \begin{matrix} n\\ m\end{matrix} \right)^{-1}\sum \phi_ N(X_{j_ 1},...,X_{j_ m})\), where the sum is over all \(1\leq j_ 1<...<j_ m\leq n\) and \(\phi_ N\) is a symmetric Borel function with \(E \phi_ N(X_ 1,...,X_ m)=0.\) The authors prove asymptotic normality as \(N\to \infty\), \(n\to \infty\) of U under conditions involving \(g(x)=E \phi_ N(X_ 1,...,X_ m| \quad X_ 1=x)\) and Var g(X\({}_ 1)\). They also prove a corresponding Berry- Esséen bound under the condition \(0<n/N\leq \lambda <1\). These theorems improve the results by the authors in Sci. Sin., Ser. A 30, No.2, 113-127 (1987; Zbl 0665.60029).

Recommendations

Cites work

Cited in

(22)

This page was built for publication: Normal approximation for finite-population U-statistics

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q757968)