Weak convergence of smoothed empirical processes: Beyond Donsker classes
Let \(X_1,\dots,X_n\) be a sequence of independent random variables with common distribution \(P\) on the real line. A smoothed empirical process \(\sqrt{n}(P_n\ast K_h-P)(f)_{f\in{\mathcal F}}\) based on a kernel estimate \(P_n\ast K_h\) is studied. This process is indexed by some class \({\mathcal F}\) of uniformly bounded functions. In the usual setting the theory of empirical processes studies the behaviour of the ordinary empirical process \(\sqrt{n}(P_n-P)\), indexed by a family of sets, or more generally, a class of uniformly bounded functions \({\mathcal F}\). A class \({\mathcal F}\) is said to be a \(P\)-Donsker class if the empirical process converges to a tight Gaussian limit. Assuming that \(P\) has a smooth density, it is shown that the smoothed empirical process converges weakly to a tight Gaussian limit for classes \({\mathcal F}\) that are not necessarily \(P\)-Donsker. Define \(P_{n,h}f=\int f(x)p_{n,h}(x) dx\), where \(p_{n,h}=(nh)^{-1}\sum_{i=1}^nK((x-X_i)/h)\), \(K(x)\) is a kernel. The following theorem is proved: Suppose that \(p\) is twice differentiable. Let \(\delta_0>0\) and \(A\subset{\mathbb R}\) such that \(p(x)>\delta_0\) for \(x\in A\), \(\sup_{x\in A}|p'(x)|<\infty\) and \(\sup_{x\in A}|p(x)|<\infty\). Let \({\mathcal F}\) be a class of uniformly bounded functions which are zero outside the set \(A\). Then a necessary and sufficient condition for the weak convergence of \(\sqrt{n}(P_{n,h}-P)(f)_{f\in{\mathcal F}}\) to \(G_P(f)_{f\in{\mathcal F}}\) is that \({\mathcal F}\) is \(P\)-pre-Gaussian. We recall that a class \({\mathcal F}\) is called \(P\)-pre-Gaussian if there exists a tight Gaussian process \(G_P\) in \(l^\infty({\mathcal F})\), which is indexed by \({\mathcal F}\). A \(P\)-Donsker class is \(P\)-pre-Gaussian, but the converse is not true. For the entire collection see [Zbl 0948.00040].
- scientific article; zbMATH DE number 720761
- scientific article; zbMATH DE number 96990
- Weak convergence of smoothed empirical process in Hölder spaces
- Necessary and sufficient conditions for weak convergence of smoothed empirical processes.
- Characterization of weak convergence for smoothed empirical and quantile processes under \(\varphi\)-mixing
- Weak convergence and empirical processes. With applications to statistics
- The central limit theorem for U-processes indexed by Hölder's functions
- Weak convergence of smoothed empirical process in Hölder spaces
- Necessary and sufficient conditions for weak convergence of smoothed empirical processes.
- A Donsker theorem for Lévy measures
- A uniform central limit theorem and efficiency for deconvolution estimators
- On smoothness measurement for weakly stationary processes
- Donsker results for the empirical process indexed by functions of locally bounded variation and applications to the smoothed empirical process
- On the optimal estimation of probability measures in weak and strong topologies
- scientific article; zbMATH DE number 96990 (Why is no real title available?)
- scientific article; zbMATH DE number 720761 (Why is no real title available?)
- On the convergence rate of potentials of Brenier maps
- Efficient prediction for linear and nonlinear autoregressive models
- Donsker-type theorems for nonparametric maximum likelihood estimators
- Uniform central limit theorems for kernel density estimators
This page was built for publication: Weak convergence of smoothed empirical processes: Beyond Donsker classes
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4524737)