Upper Bounds on Poisson Tail Probabilities.
Abstract
Let p(k) = exp(-lambda) lamda sub k/k (k= 0,1,...) be the Poisson mass function. In a variety of application contexts, it is necessary to computer infinite sums involving these probabilities. For example, such sums occur naturally in numerical algorithms developed for Poisson variate generation purposes and for computing terminal rewards of uniformizable continuous-time Markov chains. From a practical standpoint, it is necessary to truncate these infinite sums after a finite number of terms. Development of a priori error bounds on the error incurred by this kind of truncation requires bounds on the left and right tails of the Poisson distribution; such bounds are given here. These bounds are easily computable in a numerically stable way, even when the Poisson parameter lambda is large.
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 01, 1986
- Accession Number
- ADA167531
Entities
People
- Peter W. Glynn
Organizations
- University of Wisconsin–Madison