scientific article; zbMATH DE number 1337347
Let \(x=\{x(n), n\geq 1\}\) be a sequence of Gaussian random variables with \(Ex(n)=0, Ex^2(n)=\sigma^2(n)>0, r(i,j)=E(x(i)x(j))/\sigma(i)\sigma(j)\) and let \(f=\{f(n),n\geq 1\}\) be a sequence of positive numbers such that \(\lim_{n\to\infty} f(n)=\infty\). The main results of the paper are the following. If \(f(n)\leq f(n+1)\) and \({}r(i,j){}\leq c_1 f(i)/f(j), 1\leq i<j, j\geq 2\), then a.s. with \( \chi(n)=(2\sigma^2(n)V_f(n))^{1/2} \) and \( V_f(n)=\log\left(\sum_{i=1}^{n-1} \min\{C,\log(f(i+1)/f(i))\}\right) \) \[ \lim_{n\to\infty}\sup \chi^{-1}(n)x(n)\geq 1. \] If \(f(n)\leq f(n+1)\) and \(r(i,j)\geq f(i)/f(j)\) or \(r(i,j)\geq (-1)^{j-i}f(i)/f(j), 1\leq i<j, j\geq 2\), then a.s. \[ \lim_{n\to\infty}\sup \chi^{-1}(n)x(n)\leq 1. \] If \(c_2 f(n)\geq f(n+1)\) and \({}r(i,j){}\leq c_1 f(i)/f(j), 1\leq i<j, j\geq 2\), then a.s. \[ \lim_{n\to\infty}\sup (2\sigma^2(n)\log\log f(n))^{-1/2}x(n)\geq 1. \] These results are applied to investigate the asymptotic properties of solutions of stochastic difference equations.
- Bounded laws of the iterated logarithm for quadratic forms in Gaussian random variables
- The law of the iterated logarithm for non-instantaneous filters of strongly dependent Gaussian sequences
- Power type law of the iterated logarithm for Gaussian variables
- Law of the iterated logarithm for unstable Gaussian autoregressive models
- Gebelein's inequality and its consequences
- scientific article; zbMATH DE number 4034756 (Why is no real title available?)
This page was built for publication:
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4264018)
