EDGEWORTH EXPANSIONS FOR FUNCTIONALS OF U-STATISTICS
Let \(X_1\), \(X_2, \dots, X_n, \dots\) be independently and identically distributed random variables with distribution function \(F(x)\). Let \(h(x_1, \dots, x_r)\) be a real valued function which is symmetric in its arguments and let \(\theta= Eh(X_1, \dots, X_r)\). For \(n\geq r\) let us define a \(U\)-statistic by \[ U_n= {n\choose r}^{-1} \sum_{C_{n,r}} h(X_{i_1}, \dots, X_{i_r}), \] where \(\sum_{C_{n,r}}\) indicates that the summation is taken over all integers \(i_1, \dots, i_r\) satisfying \(1\leq i_1< \cdots <i_r\leq n\). Let \(f\) be a Borel measurable function which satisfies some conditions on the differentiability. Using Slutsky's theorem, we can show that \(\sqrt n[f (U_n) -f(\theta)]/[r \xi| f'(\theta) |]\) converges to the standard normal in law. \textit{M. Ghosh} [Sankyā, Ser. A 47, 255-270 (1985; Zbl 0584.62028)] obtained the Berry-Esséen bound under the conditions that the third moment of the kernel \(h\) exists and \(f\) is bounded. Ghosh also obtained the Berry-Esséen bound for a studentized functional \[ \left[ nf(U_n)- {n-1 \over n} \sum^n_{i=1} f(U_n^{(i)}) -f(\theta) \right]/ \widehat \sigma_n(f), \] \[ \text{where} \quad \widehat \sigma^2_n(f) ={n-1\over n} \sum^n_{i=1} \bigl[f(U_n^{(i)})-f(U_n)\bigr]^2. \] Ghosh proved the Berry-Esséen bound \(O(n^{-1/2})\) under the conditions that the \((4+ \varepsilon)\)th moment of the kernel exists and \(f\) is bounded. We will obtain an approximation of \(n\widehat \sigma^2_n(f)\) and derive Edgeworth expansions of the standardized and studentized functionals of \(U\)-statistics with remainder term \(o(n^{-1/2})\). In Section 2 we will prepare the notation and review previously obtained results. In Section 3 we will obtain approximations of the standardized functionals and derive Edgeworth expansions, and in Section 4 we will consider the studentized case. Finally, in Section 5 we discuss the estimation of the standard error, which is a function of the variance and does not satisfy the boundedness condition of the second derivative. For the sake of simplicity, we will consider the kernel of degree 2. The generalization to the kernel with arbitrary degree will be easily obtained except for notational complications.
- Expansions for U-statistics and von Mises functionals
- Edgeworth expansion for \(U\)-statistics under minimal conditions.
- On the rate of convergence and asymptotic expansions for \(U\)-statistics under alternatives
- The Edgeworth expansion for U-statistics of degree two
- Weighted bootstrap for \(U\)-statistics
- The Berry-Esséen bound for Studentized statistics
- An Edgeworth expansion and a normalizing transformation for L-statistics
- Around U-statistics
- scientific article; zbMATH DE number 5711156 (Why is no real title available?)
- scientific article; zbMATH DE number 3936187 (Why is no real title available?)
- scientific article; zbMATH DE number 4024401 (Why is no real title available?)
- scientific article; zbMATH DE number 4060513 (Why is no real title available?)
- scientific article; zbMATH DE number 1775004 (Why is no real title available?)
- ON THE NORMAL APPROXIMATIONS OF STUDENTIZED U-STATISTIC
- Two-term Edgeworth expansions for the classes of \(U\)- and \(V\)-statistics
- The Edgeworth Expansion for U-Statistics of Degree Two
- Edgeworth expansions for statistics which are functions of lattice and non-lattice variables
This page was built for publication: EDGEWORTH EXPANSIONS FOR FUNCTIONALS OF U-STATISTICS
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4351492)
