No label defined (Q4396539)
From MaRDI portal
!
This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:
scientific article; zbMATH DE number 1166241
| Language | Label | Description | Also known as |
|---|---|---|---|
| default for all languages | No label defined |
||
| English | No label defined |
scientific article; zbMATH DE number 1166241 |
Statements
27 February 2000
0 references
Poisson process
0 references
generalized Cox process
0 references
limit theorem
0 references
rapidity of convergence
0 references
exponential moments
0 references
0.8511819243431091
0 references
0.8404834866523743
0 references
0.8371706604957581
0 references
Let \(N_{1}\), \(\Lambda\), \(X_{1},\dots,X_{n}\),\dots be independent, where \(N_{1}\) is a Poisson process with parameter 1, \(\Lambda\) is a process with nondecreasing right continuous sample paths, \(\Lambda (0)= 0\), \(E\Lambda (t)= t\), and \(X_{i}\) are identically distributed random variables, with dispersion \(\sigma ^{2} \in (0,\infty)\) and \(a= EX_{i}\). Consider \(S(t)= X_{1}+\dots+X_{N _{1} ( \Lambda ( t))}\) (the generalized Cox process). Let \(\Lambda (t)\rightarrow \infty\) in probability for \(t\rightarrow +\infty\). The authors prove that the distribution \(F_{t}\) of \((S(t)-at)/\sigma t^{1 / 2}\) converges to \(Z\), for \(t\rightarrow \infty\), if and only if \((\Lambda (t)-t)t^{- 1 / 2}\) converges in distribution to a \(V\) such that \(Z\) is the convolution of the distribution of \(aV/\sigma\) and a centered Gaussian with dispersion \(1+a^{2} \sigma ^{- 2}\). Let \(\Phi\) be the standard Gaussian distribution function. Suppose that \(E|V|< \infty\), \(\beta= E|X_{i} |^{3} < \infty\). The authors prove an estimation of the rapidity of convergence, namely that \[ \sup_{x} |F_{t}((-\infty ,x))-E\Phi ((\sigma x-aV)(\sigma ^{2}+a^{2})^{- 1 / 2})|\leq t^{- 1 / 2}(1.5L_{3}+2.3E|V|)+\omega (t), \] for \(-t^{- 1 / 2}\) not an atom of \(V\), where we mention that \(L_{3}\) is expressed in terms of \(a\), \(\sigma\), \(\beta\), \(E(|X_{i}-a|^{3})\) and that \(\omega (t)\) is the sum of two integrals, one integrand involving the difference between the distribution functions of \((\Lambda (t)-t)t^{- 1 / 2}\) and \(V\), and the other the distribution function of \(V\) itself. If \(V= V_{0}-EV_{0}\), \(V \geq 0\), \(E\exp(hV_{0}) \leq P(h)\exp(\gamma h^{2})\) for all \(h \geq 0\) with a polynomial \(P\), if \(E|X_{i} |^{k} < \infty\), \(k \geq 3\), and if \(\limsup_{|s |L_\infty } |E(\exp(isX_{j})|< 1\), then the authors state another inequality in which under sup appears also \(-\sum _{j = 1} ^{k - 2}w_{j}(x)t^{- j / 2}\) and in the right member an \(\varepsilon (t)\), with explicit expressions for \(\varepsilon\) and \(w_{j}\) [its proof is considered as on the same line as one figuring in the book of \textit{B. V. Gnedenko} and \textit{V. Yu. Korolev}, ``Random summation. Limit theorems and applications'' (1996; Zbl 0857.60002)]. Examples of \(\Lambda\) are given. In the case with discrete time, \(\Lambda (n)= Z_{1}+\dots+Z_{n}\) with i.i.d. \(Z_{i} \geq 0\) and \(E\exp(hZ_{i}) \leq P(h)\exp(\gamma h)\), an analogous inequality is also stated.
0 references