Error Bounds for Exponential Approximations to Geometric Convolutions.
Abstract
This paper defines Y sub 0 to be a geometric convolution of X if Y sub 0 is the sum of N sub 0 i.i.d. random variables distributed as X, where N sub 0 is geometrically distributed and independent of X. It is known that if X is non-negative with finite second moment then as p approaches limit of 0, Y sub 0/EY sub 0 converges in distribution to an exponential distribution with mean 1. Derive is an upper bound for d(Y sub 0), the distance between Y sub 0 and an exponential with mean Y sub 0, namely for 0<p < or = 1/2, d(sub 0) < or = cp where c = sq ex/sq (ex). This bound is asymptotically (p approaches limit of 0) tight.
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 15, 1986
- Accession Number
- ADA185480
Entities
People
- Mark O. Brown
Organizations
- City College of New York