Regenerative Simulation for Extreme Values.
Abstract
Let (X sub t : t > or = 0) denote the regenerative process being simulated and assume that X sub t converges weakly (in distribution) to a limit random variable X. Our concern here is in estimating the extreme values of the process (X sub t : > or = 0). Suppose we are interested in the largest value attained in the interval (0, t): (X sub t)* = sup (X sub s : 0 < or = s < or = t). Examples of this are the maximum queue lengths or waiting times in a queueing system. As t increases so will (X sub t)*, without bound if the state space of (X sub t : > or = 0) is unbounded. This report developes two methods for estimating the distribution of (X sub t)*. When the regenerative process is either the GI/G/1 queue or a birth-death process theoretical results are available for the distribution of (X sub t)*. The waiting time, queue length, and virtual waiting time for an M/M/1 queue were simulated. The two methods for estimating the distribution of (X sub t)* were employed and the simulation results compared with the theoretical results. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 1977
- Accession Number
- ADA047945
Entities
People
- Donald Iglehart
Organizations
- Stanford University