On convergence in distribution of empirical processes defined by independent random processes (Q1925160): Difference between revisions
From MaRDI portal
Changed an Item |
ReferenceBot (talk | contribs) Changed an Item |
||
(One intermediate revision by one other user not shown) | |||
Property / MaRDI profile type | |||
Property / MaRDI profile type: MaRDI publication profile / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Q5560061 / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: On the central limit theorem in \(D[0,1]\) / rank | |||
Normal rank |
Latest revision as of 15:27, 24 May 2024
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | On convergence in distribution of empirical processes defined by independent random processes |
scientific article |
Statements
On convergence in distribution of empirical processes defined by independent random processes (English)
0 references
15 December 1996
0 references
Let \((\Omega, {\mathfrak A}, {\mathbf P})\) be a probability space and \((T, {\mathfrak B})\) be a measurable space. Denote by \(X(t)\), \(t\in T\), a measurable process defined on \((\Omega, {\mathfrak A}, {\mathbf P})\). Let \(G(t)(x)\), \(x\in{\mathbf R}\), be distribution functions of random variables \(X(t)\), \(t\in T\), and assume that there exists a bounded increasing function \(m: {\mathbf R} \to {\mathbf R}\) such that for all \(t\in T\), \(x,y\in {\mathbf R}\), \(x\leq y\), \(G(t)(y) -G(t)(x) \leq M(y)- M(x)\). Let \(\xi\), \(\xi_n\), \(n\in{\mathbf N}\), be i.i.d. random variables with values in \((T, {\mathfrak B})\) and defined on another probability space \((\Omega_1, {\mathfrak A}_1, {\mathbf P}_1)\). Define the centered empirical processes \[ F_n(x,\omega,\omega_1) =n^{-1} \sum^n_{i=1} \biggl(I_{\{X_i(\xi_i(\omega_1)) \leq x\}} (\omega)- G\bigl(\xi_i(\omega_1) \bigr) (x)\biggr), \] \(x\in {\mathbf R}\), \(\omega\in\Omega\), and \(\omega_1\in\Omega_1\). Denote by \(D_0({\mathbf R})\) the complete metric space of bounded functions \(g:{\mathbf R} \to {\mathbf R}\), continuous from the right, having limits from the left and vanishing at infinity, which is a natural generalization of the Skorokhod space \(D[0,1]\). Let \(Y_n(\omega_1) =Y_n(x,\omega, \omega_1)\), \(n\in {\mathbf N}\), be random elements in \(D_0 ({\mathbf R})\), defined by the random processes \(n^{1/2} F_n(x, \omega, \omega_1)\). The author proves that, if \(M\) is continuous from the right, then \({\mathcal L} (Y_n(\omega_1)) \to\gamma^0\) as \(n\to\infty\) weakly in \(D_0 ({\mathbf R})\) for almost all \(\omega_1\in \Omega_1\). Here \(\gamma^0\) denotes the Gaussian distribution on \(D_0({\mathbf R})\) with the same finite-dimensional distributions as the Gaussian process \(W^0_1\) satisfying for all \(x\in {\mathbf R}\), \({\mathbf E} W^0_1(x) =0\), and for all \(x,y\in {\mathbf R}\), \(x\leq y\), \[ {\mathbf E} W^0_1(x) W^0_1(y) ={\mathbf E}_1 \biggl(G\bigl(\xi (\omega_1)\bigr) (x)\biggr) \biggl(1-G \bigl( \xi (\omega_1) \bigr) (y)\biggr), \] where \({\mathbf E}_1\) is the expectation with respect to \({\mathbf P}_1\). The author also proves this convergence for the space of continuous bounded functions vanishing at infinity.
0 references
empirical processes
0 references
convergence in distribution
0 references
Gaussian random processes
0 references