Compound Poisson Approximation of the Number of Exceedances in Gaussian Sequences |
| |
Authors: | Mikael Raab |
| |
Institution: | (1) Department of Mathematics, Royal Institute of Technology, S-100 44 Stockholm, Sweden |
| |
Abstract: | Consider a finite sequence of Gaussian random variables. Count the number of exceedances of some level a, i.e. the number of values exceeding the level. Let this level and the length of the sequence increase simultaneously so that the expected number of exceedances remains fixed. It is well-known that if the long-range dependence is not too strong, the number of exceeding points converges in distribution to a Poisson distribution. However, for sequences with some individual large correlations, the Poisson convergence is slow due to clumping. Using Steins method we show that, at least for m-dependent sequences, the rate of convergence is improved by using compound Poisson as approximating distribution. An explicit bound for the convergence rate is derived for the compound Poisson approximation, and also for a subclass of the compound Poisson distribution, where only clumps of size two are considered. Results from numerical calculations and simulations are also presented. |
| |
Keywords: | convergence rates compound Poisson approximation Stein-Chen method extreme values Gaussian sequences |
本文献已被 SpringerLink 等数据库收录! |
|