Computing Reviews
Today's Issue Hot Topics Search Browse Recommended My Account Log In
Review Help
Search
Simulation modeling and analysis (5th ed.)
Law A., McGraw-Hill Education Group, New York, NY, 2015. Type: Book (9780073401324)
Date Reviewed: Jun 21 2017

Simulation models imitate the behavior of complex systems, including manufacturing processes and real-world service facilities. They are designed to gain some understanding of how the actual system behaves under different circumstances, to evaluate the impact of design decisions, or just to predict how the system under simulation will respond to changes in its environment. Only the simplest models can be analytically studied using mathematical techniques, so computer simulation is required. Computer simulations provide numerical results. Results must be properly analyzed to avoid misleading conclusions and support decision making. This is the main focus of Law’s thorough monograph on simulation. The book is more than 700 pages long, yet it is still just an introduction, with special emphasis on the proper way to conduct a simulation study.

The first 200 pages introduce the basic concepts. In particular, they delve into the implementation of discrete-event simulation programs. The first chapter provides two fairly basic implementations in ANSI C: a single-server queueing system and an inventory system. The second chapter revisits the queueing system and provides three more complex examples: a time-sharing system, a multi-teller bank with jockeying (customers moving from one queue to another), and a job-shop scheduling model. The examples are coded on top of a simple simulation library to streamline their implementation in C. The author’s library, called simlib, is provided with its complete source code (included as an appendix and downloadable). The third chapter in this introductory section describes the desirable features of simulation software and provides the same example implemented using three different simulation software packages (Arena, ExtendSim, and Simio). Once the basics are covered in detail, the author provides a short review of probability and statistics concepts: random variables; the estimation of means, variances, and correlations; and confidence intervals, including the common ones based on the student t-test and the less common Willink’s confidence intervals for asymmetric distributions [1].

The really interesting content starts after those introductory chapters. Law starts by describing how to build valid, credible, and appropriately detailed simulation models. His advice on increasing model validity and model credibility is invaluable for practitioners. Law describes how to find real-world observations and compare them with simulation outputs following statistically sound procedures. After that, he delves into choosing input probability distributions, given that the simulation model is typically driven by randomly generated input data. He analyzes many common and not so common probability distributions. Selecting the right distribution family, the right distribution, and estimating its parameters are common problems that simulation designers must solve for their models to be useful.

Almost 100 pages are devoted to generating random samples from probability distributions, mainly from linear congruential generators (LCGs) and combined multiple recursive generators (MRGs). C code for two of them is also provided: a simple prime modulus multiplicative LCG (PMMLCG) and a more robust L’Ecuyer’s MRG. Apart from individual pseudorandom number generators, the author also discusses different approaches for generating random variates (for example, inverse transform, composition, or rejection) and provides details for generating random variates for all the distribution probabilities one might consider for modeling simulation inputs. After his detailed and informative chapters on input modeling, Law turns his attention toward the proper statistical analysis of simulation output data. He dissects the transient and steady-state behavior of simulated systems. Simulation driven by random inputs produces random outputs that must be properly analyzed from a statistical point of view, so he delves into how to compute confidence intervals for the output, from a single system simulation (chapter 9) and from different simulation models (chapter 10), which might involve the application of the Bonferroni correction. When those confidence intervals are too wide, five variance reduction techniques can help improve the precision of the output (chapter 11).

Law also devotes a couple of chapters to experimental design issues. He describes how to design simulation experiments using 2k factorial design and 2k-p fractional factorial design. He concludes with a discussion on metamodels and simulation-based optimization. Two additional chapters conclude this volume. The first one briefly mentions agent-based simulation as a particular case of discrete-event simulation and system dynamics as a kind of continuous simulation (using Lokta-Volterra differential equations for predator-prey systems as a recurring example). Some final notes on Monte Carlo and spreadsheet simulation are also included in this short chapter, which could deserve a whole book on its own. The second chapter, available online, provides a more conventional perspective on the simulation of manufacturing systems, a key application area.

Overall, the book is thorough, detailed enough to be self-contained, and tiresome at times. Fortunately, the author has interspersed dozens of real-world examples within the text, which help the reader become engaged. The problems at the end of each chapter enhance its adoption as an undergraduate textbook on simulation. The book is also a good starting point for self-study for those who are really motivated to study statistically sound computer simulation.

Reviewer:  Fernando Berzal Review #: CR145361 (1708-0498)
1) Willink, R. A confidence interval and test for the mean of an asymmetric distribution. Communications in Statistics - Theory and Methods 34, 4(2005), 753–766.
Bookmark and Share
  Featured Reviewer  
 
Modeling And Prediction (D.4.8 ... )
 
 
Simulation (D.4.8 ... )
 
Would you recommend this review?
yes
no
Other reviews under "Modeling And Prediction": Date
A model for the stability analysis of maintenance strategies for linear list
Bastani F., Chen I., Hilal W. The Computer Journal 34(1): 80-87, 1991. Type: Article
Feb 1 1992
Disk performance in a transaction-oriented system
Heyman D., Tsur S. SIAM Journal on Computing 13(4): 669-681, 1984. Type: Article
Aug 1 1985
Response times in level-structured systems
Paul K. J. ACM Transactions on Computer Systems 5(3): 232-248, 1987. Type: Article
Jul 1 1988
more...

E-Mail This Printer-Friendly
Send Your Comments
Contact Us
Reproduction in whole or in part without permission is prohibited.   Copyright 1999-2024 ThinkLoud®
Terms of Use
| Privacy Policy