The purpose of this report is to provide an introduction to the regenerative method for simulation analysis. The simulations are simulations of stochastic systems, i.e., systems with random elements. The regenerative approach leads to a statistical methodology for analyzing the output of those simulations which have the property of 'starting afresh probabilistically' from time to time. The class of such simulations is very large and very important, including simulations of a broad variety of queues and queueing networks, inventory systems, inspection, maintenance, and repair operations, and numerous other situations.
A wide variety of stochastic systems may be viewed as Markov chains taking on values in a general state space. An example is the class of generalized semi-Markov processes, which are commonly obtained in network queueing problems via the technique of supplementary variables. A simulator is often interested in obtaining steady state properties of such a system. Some recent developments in Markov chain theory by Athreya, Ney, and Nummelin allow one to embed a certain subclass of these processes in a regenerative environment. We study some consequences of this embedding and develop statistical estimation procedures for the general problem that bear close resemblance to the regenerative method of simulation analysis for finite state Markov chains. (Author).
Students of statistics, operations research, and engineering will be informed of simulation methodology for problems in both mathematical statistics and systems simulation. This discussion presents many of the necessary statistical and graphical techniques. A discussion of statistical methods based on graphical techniques and exploratory data is among the highlights of Simulation Methodology for Statisticians, Operations Analysts, and Engineers. For students who only have a minimal background in statistics and probability theory, the first five chapters provide an introduction to simulation.
Simulation is a controlled statistical sampling technique that can be used to study complex stochastic systems when analytic and/or numerical techniques do not suffice. The focus of this book is on simulations of discrete-event stochastic systems; namely, simulations in which stochastic state transitions occur only at an increasing sequence of random times. The discussion emphasizes simulations on a finite or countably infinite state space. * Develops probabilistic methods for simulation of discrete-event stochastic systems * Emphasizes stochastic modeling and estimation procedures based on limit theorems for regenerative stochastic processes * Includes engineering applications of discrete-even simulation to computer, communication, manufacturing, and transportation systems * Focuses on simulations with an underlying stochastic process that can specified as a generalized semi-Markov process * Unique approach to simulation, with heavy emphasis on stochastic modeling * Includes engineering applications for computer, communication, manufacturing, and transportation systems
Simulation means driving a model of a system with suitable inputs and observing the corresponding outputs. It is widely applied in engineering, in business, and in the physical and social sciences. Simulation method ology araws on computer. science, statistics, and operations research and is now sufficiently developed and coherent to be called a discipline in its own right. A course in simulation is an essential part of any operations re search or computer science program. A large fraction of applied work in these fields involves simulation; the techniques of simulation, as tools, are as fundamental as those of linear programming or compiler construction, for example. Simulation sometimes appears deceptively easy, but perusal of this book will reveal unexpected depths. Many simulation studies are statistically defective and many simulation programs are inefficient. We hope that our book will help to remedy this situation. It is intended to teach how to simulate effectively. A simulation project has three crucial components, each of which must always be tackled: (1) data gathering, model building, and validation; (2) statistical design and estimation; (3) programming and implementation. Generation of random numbers (Chapters 5 and 6) pervades simulation, but unlike the three components above, random number generators need not be constructed from scratch for each project. Usually random number packages are available. That is one reason why the chapters on random numbers, which contain mainly reference material, follow the ch!lPters deal ing with experimental design and output analysis.
