Bayesian Analysis of Stochastic Process Models

9

Discrete event simulation

9.1 Introduction

Typically, once an organization has realized that a system is not operating as desired, it will look for ways to improve its performance. Sometimes it will be possible to experiment with the real system and, through observation and the aid of statistical techniques, reach valid conclusions to better the system. However, experiments with a real system may entail ethical and/or economical problems, which may be avoided by dealing with a prototype, that is, a physical model of the system. Sometimes it is not feasible to build such a prototype, and as an alternative, we may be able to develop a mathematical model that captures the essential behavior of the system. This analysis may sometimes be carried out through analytical or numerical methods. However, in other cases, the model may be too complex to be dealt with in such a way. In such extreme cases, we may use simulation. Large, complex, system simulation has become common practice in many industrial and service areas such as the performance prediction of integrated circuits, the behavior of controlled nuclear fusion devices, or the performance evaluation of call centers.

In this chapter, we shall focus on discrete event simulation (DES), which refers to computer based experimentation with a system which stochastically evolves in time, but which cannot be easily analyzed via standard numerical methods, including Markov chain Monte Carlo (MCMC) methods. This system might already exist, ...