Proceedings of the 2001 Winter Simulation Conference

B. A. Peters, J. S. Smith, D. J. Medeiros, and M. W. Rohrer, eds.

ABSTRACT
In this tutorial we present techniques for building valid and
credible simulation models. Ideas to be discussed include
the importance of a definitive problem formulation, discus-
sions with subject-matter experts, interacting with the deci-
sion-maker on a regular basis, development of a written
conceptual model, structured walk-through of the concep-
tual model, use of sensitivity analysis to determine impor-
tant model factors, and comparison of model and system
performance measures for an existing system (if any).
Each idea will be illustrated by one or more real-world ex-
amples. We will also discuss the difficulty in using formal
statistical techniques (e.g., confidence intervals) to validate
simulation models.
1 WHAT IS MODEL VALIDATION
Use of a simulation model is a surrogate for experimenta-
tion with the actual system (existing or proposed), which is
usually disruptive, not cost-effective, or simply impossible.
Thus, if the model is not a close approximation to the ac-
tual system, any conclusions derived from the model are
likely to be erroneous and may result in costly decisions
being made. Validation should and can be done for all
models, regardless of whether the corresponding system
exists in some form or whether it will be built in the future.
We now give definitions of validation and credibility.
Validation is the process of determining whether a simula-
tion model is an accurate representation of the system, for
the particular objectives of the study. The following are
some general perspectives on validation:
Conceptually, if a simulation model is valid,
then it can be used to make decisions about the
system similar to those that would be made if it
were feasible and cost-effective to experiment
with the system itself.
The ease or difficulty of the validation process de-
pends on the complexity of the system being mod-
eled and on whether a version of the system cur-
rently exists (see Section 2.8). For example, a
model of a neighborhood bank would be relatively
easy to validate since it could be closely observed.
On the other hand, a model of the effectiveness of a
naval weapons system in the year 2025 would be
virtually impossible to validate completely, since
the location of the battle and the nature of the en-
emy weapons would be unknown. Also, it is often
possible to collect data on an existing system that
can be used for building and validating a model.
A simulation model of a complex system can only
be an approximation to the actual system, no mat-
ter how much time and money is spent on model
building. There is no such thing as absolute
model validity, nor is it even desired. Indeed, a
model is supposed to be an abstraction and simpli-
fication of reality. The more time (and hence
money) that is spent on model development, the
more valid the model should be in general. How-
ever, the most valid model is not necessarily the
most cost-effective. For example, increasing the
validity of a model beyond a certain level might
be quite expensive, since extensive data collection
may be required, but might not lead to signifi-
cantly better insight or decisions.
A simulation model should always be developed
for a particular set of objectives. In fact, a model
that is valid for one objective may not be for an-
other.
Validation is not something to be attempted after
the simulation model has already been developed,
and only if there is time and money remaining.
Unfortunately, our experience indicates that this
recommendation is often not followed.

Example 1. An organization paid a consulting com-
pany $500,000 to perform a simulation study that
had a six-month duration. After the study was sup-
posedly completed, a person from the client organiza-



HOW TO BUILD VALID AND CREDIBLE SIMULATION MODELS


Averill M. Law
Michael G. McComas
Averill M. Law and Associates, Inc.
P.O. Box 40996
Tucson, AZ 85717, U.S.A.



22
Law and McComas

tion called us and asked, Can you tell me in five min-
utes on the phone how to validate our model?

A simulation model and its results have credibility if
the decision-maker and other key project personnel accept
them as correct. Note that a credible model is not neces-
sarily valid, and vice versa. The following things help es-
tablish credibility for a model:

The decision-makers understanding and agree-
ment with the models assumptions
Demonstration that the model has been validated
and verified (i.e., that the model computer pro-
gram has been debugged)
The decision-makers ownership of and involve-
ment with the project
Reputation of the model developers

Note that many of the ideas and examples presented in
this paper are based on the chapter Building Valid, Credi-
ble, and Appropriately Detailed Models in Law and Kelton
(2000) and also on the simulation short courses presented by
the first author. Other references on model validation are
Balci (1998), Banks, Carson, Nelson, and Nicol (2001), Car-
son (1986), Sargent (1996), and Shannon (1975).
The remainder of this paper is organized as follows.
We present in Section 1.1 a seven-step approach for con-
ducting a successful simulate study. In Section 2 we dis-
cuss techniques for developing a more valid and credible
simulation model. Guidelines for obtaining good model
data are given in Section 3. Finally, Section 4 provides a
summary of the most important validation ideas.
1.1 A Seven-Step Approach for Conducting
a Successful Simulation Study
In Figure 1 we present a seven-step approach for conduct-
ing a successful simulation study. Having a definitive ap-
proach for conducting a simulation study is critical to the
studys success in general and to developing a valid model
in particular. In Section 2 we will relate each of our vali-
dation/credibility enhancement techniques to one or more
of these steps.
We now discuss important activities that take place in
each of the seven steps.
1.1.1 Step 1. Formulate the Problem
Problem of interest is stated by the decision-maker
A kickoff meeting(s) for the simulation project is (are)
conducted, with the project manager, the simulation
analysts, and subject-matter experts (SMEs) in atten
dance. The following things are discussed at this
meeting:

The overall objectives of the study
The specific questions to be answered by the
study (without such specificity it is impossible to
determine the appropriate level of model detail)
The performance measures that will be used to
evaluate the efficacy of different system configu-
rations
The scope of the model
The system configurations to be modeled
The time frame for the study and the required re-
sources





































Figure 1: A Seven-Step Approach for Conducting a Suc-
cessful Simulation Study
6
7
Program the Model
Collect
Information/Data
and Construct
Conceptual Model
Is the Conceptual
Model Valid?
Formulate the Problem
Is the Programmed Model
Valid?
No
No
3
5
Design, Conduct,
and Analyze
Experiments
Document and Present
the Simulation Results
4
2
1
Yes
Yes
23
Law and McComas

1.1.2 Step 2. Collect Information/Data and
Construct a Conceptual Model
Collect information on the system layout and operat-
ing procedures.
Collect data to specify model parameters and probabil-
ity distributions (e.g., for the time to failure and the
time to repair of a machine).
Document the model assumptions, algorithms, and
data summaries in a written conceptual model.
The level of model detail should depend on the follow-
ing:

Project objectives
Performance measures of interest
Data availability
Credibility concerns
Computer constraints
Opinions of SMEs
Time and money constraints

There should not be a one-to-one correspondence be-
tween the model and the system.
Collect performance data from the existing system (if
any) to use for model validation in Step 5.
1.1.3 Step 3. Is the Conceptual Model Valid?
Perform a structured walk-through of the conceptual
model before an audience that includes the project
manager, analysts, and SMEs. This is called concep-
tual-model validation.
If errors or omissions are discovered in the conceptual
model, which is almost always the case, then the con-
ceptual model must be updated before proceeding to
programming in Step 4.
1.1.4 Step 4. Program the Model
Program the conceptual model in either a commercial
simulation-software product or in a general-purpose
programming language (e.g., C or C++).
Verify (debug) the computer program.
1.1.5 Step 5. Is the Programmed Model Valid?
If there is an existing system, then compare model per-
formance measures with the comparable performance
measures collected from the actual system (see Step
2). This is called results validation.
Regardless of whether there is an existing system, the
simulation analysts and SMEs should review the simu-
lation results for reasonableness. If the results are
consistent with how they perceive the system should
operate, then the simulation model is said to have face
validity.
Sensitivity analyses should be performed on the pro-
grammed model to see which model factors have the
greatest effect on the performance measures and, thus,
have to be modeled carefully.
1.1.6 Step 6. Design, Make, and Analyze
Simulation Experiments
For each system configuration of interest, decide on
tactical issues such as run length, warmup period, and
the number of independent model replications.
Analyze the results and decide if additional experi-
ments are required.
1.1.7 Step 7. Document and Present
the Simulation Results
The documentation for the model (and the associated
simulation study) should include the conceptual model
(critical for future reuse of the model), a detailed de-
scription of the computer program, and the results of
the current study.
The final presentation for the simulation study should
include animations and a discussion of the model build-
ing/validation process to promote model credibility.
2 TECHNIQUES FOR DEVELOPING
VALID AND CREDIBLE MODELS
In this section we present practical techniques for develop-
ing valid and credible models. At the end of each subsection
title, we state in square brackets ([ ]) in which of the seven
steps (at a minimum) the technique should be applied.
2.1 Formulating the Problem Precisely [1]
It is critical to formulate the problem of interest in a pre-
cise manner. This should include an overall statement of
the problem to be solved, a list of the specific questions
that the model is to answer, and the performance measures
that will be used to evaluate the efficacy of particular sys-
tem configurations. Without a definitive statement of the
specific questions of interest, it is impossible to decide on
an appropriate level of model detail. Performance meas-
ures must also be clearly stated since different measures
may dictate different levels of model detail [see Law and
Kelton (2000, pp. 678-679) for an example].
When the decision-maker first initiates a simulation
study, the exact problem to be solved is sometimes not
precisely stated or even completely understood. Thus, as
the study proceeds and a better understanding is obtained,
this information should be communicated to the decision-
maker who may reformulate the problem.
24
Law and McComas

2.2 Interviewing Subject-Matter Experts [1, 2]
There will never be a single person that knows all of the in-
formation necessary to build a simulation model. Thus, it
will be necessary for the simulation analysts to talk to many
different SMEs to gain a complete understanding of the sys-
tem to be modeled. Note that some of the information sup-
plied by the SMEs will invariably be incorrect if a certain
part of the system is particularly important, then at least two
SMEs should be queried. In Section 2.6, we will discuss a
technique that helps ensure that a models assumptions are
correct and complete this technique is also useful for re-
solving differences of opinion among SMEs.
2.3 Interacting with the Decision-Maker
on a Regular Basis [1-7]
One of the most important ideas for developing a valid and
credible model is for the analyst to interact with the deci-
sion-maker and other members of the project team on a
regular basis. This approach has the following key benefits:

Helps ensure that the correct problem is solved
The exact nature of the problem may not be
initially known.
The decision-maker may change his/her ob-
jectives during the course of the study.
The decision-makers interest and involvement in
the study are maintained.
The model is more credible because the decision-
maker understands and agrees with the models
assumptions.

Example 2. A military analyst worked on a simulation
project for several months without interacting with the
general who requested it. At the final Pentagon briefing
for the study, the general walked out after five minutes
stating, Thats not the problem Im interested in.
2.4 Using Quantitative Techniques to
Validate Components of the Model [2]
The simulation analyst should use quantitative techniques
whenever possible to test the validity of various compo-
nents of the overall model. We now give examples of
techniques that have been used for this purpose.
If one has fit a theoretical probability distribution (e.g.,
exponential or normal) to a set of observed data, then the
adequacy of the representation can be assessed by using
graphical plots and goodness-of-fit tests [see Law and Kel-
ton (2000, Chapter 6)].
As will be discussed in Section 3, it is important to use
appropriate data in building a model; however, it is equally
important to exercise care when structuring these data. For
example, if several sets of data have been observed for the
same random phenomenon, then the correctness of merg-
ing these data sets can be assessed by using the Kruskal-
Wallis test of homogeneity of populations [see Law and
Kelton (2000, pp. 394-395)]. If the data sets appear to be
homogeneous, they can be merged and the combined data
set used for some purpose in the simulation model.

Example 3. Consider a manufacturing system for
which time-to-failure and time-to-repair data were col-
lected for two identical machines made by the same
vendor. However, the Kruskal-Wallis test showed that
the two distributions were, in fact, different for the two
machines. Thus, each machine was given its own
time-to-failure and time-to-repair distributions in the
simulation model.
2.5 Documenting the Conceptual Model [2]
Communication errors are a major reason why simulation
models very often contain invalid assumptions. Docu-
menting all assumptions, algorithms, and data summaries
can lessen this problem. This report is the major documen-
tation for the model and should be readable by analysts,
SMEs, and decision-makers alike. The following are some
of the things that should be included in the conceptual
model:

An overview section that discusses overall project
goals, specific issues to be addressed by the
model, and relevant performance measures
A diagram showing the layout for the system
Detailed descriptions of each subsystem (in bullet
format for easy reading) and how they interact
What simplifying assumptions were made and
why
Summaries of model input data (technical analy-
ses should be put in appendices to promote report
readability by decision-makers)
Sources of important or controversial information

The conceptual model should contain enough detail so
that it is a blueprint for creating the simulation computer
program (in Step 4).
2.6 Performing a Structured Walk-Through
of the Conceptual Model [3]
As previously discussed, the simulation analyst will need
to collect system information from many different SMEs.
Furthermore, these people are typically very busy dealing
with the daily problems that occur within their organiza-
tion, often resulting in their giving something less than
their undivided attention to the questions posed by the
simulation analyst. As a result, there is a considerable
danger that the analyst will not obtain a complete and cor-
25
Law and McComas

rect description of the system. An effective way of dealing
with this potential problem is to conduct a structured walk-
through of the conceptual model before an audience of
SMEs and decision-makers. Using a projection device, the
analyst goes through the conceptual model bullet-by-bullet,
but not proceeding from one bullet to the next until every-
body in the room is convinced that a particular bullet is
correct and at an appropriate level of detail. A structured
walk-through will increase both the validity and credibility
of the simulation model. (As stated above, this exercise is
called conceptual-model validation.)
The structured walk-through should ideally be held at a
remote site (e.g., a hotel meeting room), so that people give
the meeting their full attention. Furthermore, it should be
held prior to the beginning of programming in case major
problems are uncovered at the meeting. The conceptual
model should be sent to participants prior to the meeting and
their comments requested. We do not, however, consider
this a substitute for the structured walk-through itself, since
people may not the have the time or motivation to review the
document carefully on their own. Furthermore, the interac-
tions that take place at the actual meeting are invaluable.

Example 4. At a structured walk-through of a trans-
portation system, a significant percentage of the as-
sumptions given to us by our corporate sponsor were
found to be wrong by the SMEs present. (Due to the
large geographic distances between the home offices
of the sponsor and the SMEs, it was not possible for
the SMEs to be present at the kickoff meeting for the
project.) As a result, various people were assigned re-
sponsibilities to collect information on different parts
of the system. The collected information was used to
update the conceptual model, and a second walk-
through was successfully performed.
2.7 Performing Sensitivity Analyses to Determine
Important Model Factors [5]
An important technique for determining which model fac-
tors have a significant impact on the desired measures of
performance is sensitivity analysis. If a particular factor
appears to be important, then it needs to be modeled care-
fully. The following are examples of factors that could be
investigated by a sensitivity analysis:

The value of a parameter (see Example 5)
The choice of a probability distribution
The entity moving through the simulated system
The level of detail for a subsystem

Example 5. In a simulation study of a new system,
suppose that the value of a probability is estimated to
be 0.75 as a result of conversations with SMEs. The
importance of getting the value of this probability ex-
actly correct can be determined by running the simu-
lation with 0.75 and, for example, by running it with
each of the values 0.70 and 0.80. If the three simula-
tion runs produce approximately the same results, then
the output is not sensitive to the choice of the
probability over the range 0.70 to 0.80. Otherwise, a
better specification of the probability is needed.
(Strictly speaking, to determine the effect of the
probability on the models results, we should make
several independent replications of the model using
different random numbers for each of the three cases.)

If one is trying to determine the sensitivity of the
simulation output to changes in two or more factors of in-
terest, then it is not correct, in general, to vary one factor at
a time while setting the other factors at some arbitrary val-
ues. A more correct approach is to use statistical experi-
mental design, which is discussed in Law and Kelton
(2000, Chapter 12) and in Montgomery (2000). The effect
of each factor can be formally estimated and, if the number
of factors is not too large, interactions between factors can
also be detected.
2.8 Validating the Output from the
Overall Simulation Model [5]
The most definitive test of a simulation models validity is
establishing that its output data closely resemble the output
data that would be observed from the actual system. If a
system similar to the proposed one now exists, then a simu-
lation model of the existing system is developed and its out-
put data are compared to those from the existing system it-
self. If the two sets of data compare closely, then the
model of the existing system is considered valid. (The ac-
curacy required from the model will depend on its intended
use and the utility function of the decision-maker.) The
model is then modified so that it represents the proposed
system. The greater the commonality between the existing
and proposed systems, the greater our confidence in the
model of the proposed system. There is no completely de-
finitive approach for validating the model of the proposed
system. If there were, then there might be no need for a
simulation model in the first place. If the above comparison
is successful, then it has the additional benefit of providing
credibility for the use of simulation. (As stated above, the
idea of comparing the model and system output data for the
existing system is called results validation.)

Example 6. A U.S. Air Force test agency performed a
simulation study for a bomb wing of bombers using
the Logistics Composite Model (LCOM). The ulti-
mate goal of the study was to evaluate the effect of
various proposed logistics policies on the availability
of the bombers, i.e., the proportion of time that the
bombers were available to fly missions. Data were
26
Law and McComas

available from the actual operations of the bomb wing
over a 9-month period, and included both failure data
for various aircraft components and a bomb-wing
availability of 0.9. To validate the model, the Air
Force first simulated the 9-month period with the ex-
isting logistics policy and obtained a model availabil-
ity of 0.873, which is 3 percent different than the his-
torical availability. This difference was considered
acceptable because an availability of 0.873 would still
allow enough bombers to be available for the Air
Force to meet its mission requirements.

Example 7. A manufacturer of heat-treated aluminum
products was thinking of replacing its existing batch
furnace by a new continuous furnace in order to in-
crease its production capacity [see Law (1991)]. We
first simulated the existing system and found that the
model monthly throughput differed from the historical
monthly throughput by less than one percent. Thus, it
appeared that the model of the existing system was
reasonably valid.

A number of statistical tests (t, Mann-Whitney, etc.)
have been suggested in the validation literature for compar-
ing the output data from a simulation model with those
from the corresponding real-world system [see, for exam-
ple, Shannon (1975, p. 208)]. However, the comparison is
not as simple as it might appear, since the output processes
of almost all real-world systems and simulations are non-
stationary (the distributions of the successive observations
change over time) and autocorrelated (the observations in
the process are correlated with each other). Thus, classical
statistical tests based on independent, identically distrib-
uted (IID) observations are not directly applicable. Fur-
thermore, we question whether hypothesis tests, as com-
pared with constructing confidence intervals for
differences, are even the appropriate statistical approach.
Since the model is only an approximation to the actual sys-
tem, a null hypothesis that the system and model are the
same is clearly false. We believe that it is more useful to
ask whether or not the differences between the system and
the model are significant enough to affect any conclusions
derived from the model. For a brief discussion of statisti-
cal procedures that can be used to compare model and sys-
tem output data, see Section 2.10.
Whether or not there is an existing system, analysts
and SMEs should review simulation output (numerical re-
sults, animations, etc.) for reasonableness. (Care must be
taken in performing this exercise, since if one knew exactly
what output to expect, then there would be no need for a
model.) If the simulation results are consistent with per-
ceived system behavior, then, as stated above, the model is
said to have face validity.

Example 8. The above idea was put to good use in
the development of a simulation model of the U.S.
Air Force manpower and personnel system. (This
model was designed to provide Air Force policy ana-
lysts with a system-wide view of the effects of vari-
ous proposed personnel policies.) The model was
run under the baseline personnel policy, and the re-
sults were shown to Air Force analysts and decision-
makers, who subsequently identified some discrep-
ancies between the model and perceived system be-
havior. This information was used to improve the
model, and after several additional evaluations and
improvements, a model was obtained that appeared
to approximate current Air Force policy closely.
This exercise improved not only the validity of the
model, but also its credibility.
2.9 Using Graphical Plots and Animations
of the Simulation Output Data [5-7]
Graphical plots (static or dynamic) and animations (dy-
namic) are useful for showing that a simulation model is
not valid and for promoting model credibility. The follow-
ing are some examples of graphical plots:

Histogram (a graphical estimate of the underlying
probability density or mass function)
Correlation plot (shows if the output data are
autocorrelated)
Time plot (one or more model variables are plot-
ted over the length of the simulation run to show
the long-run dynamic behavior of the system)
Bar charts and pie charts

An animation, which shows the short-term dynamic
behavior of a system, is useful for communicating the es-
sence of a model to decision-makers and other people who
do not understand or care about the technical details of the
model. Thus, it is a great way to enhance the credibility of
a model. Animations are also useful for verification of the
simulation computer program, for suggesting improved
operational procedures, and for training.
2.10 Statistical Techniques for Comparing
Model and System Output Data [5]
In this section we discuss the possible use of statistical
procedures for carrying out the comparison of model and
system output data discussed in Section 2.8.
Suppose that R
1
,

R
2
, , R
k
are observations from a
real-world system and that M
1
, M
2
, , M
l
are output data
from a corresponding simulation model (see Example 9 be-
low). We would like to compare these data sets in some
way to determine whether the model is an accurate repre-
sentation of the real-world system. However, most classi-
27
Law and McComas

cal statistical approaches such as confidence intervals and
hypothesis tests assume that the real-world data and the
model data are each IID data sets, which is generally not
the case (see the discussion in Section 2.8). Thus, these
classical statistical approaches are not directly applicable
to our comparison problem.

Example 9. Consider a manufacturing system where
the data of interest are the times in system of succes-
sively completed parts. These data are not independ-
ent for the actual system (nor for a corresponding
simulation model). For example, if the system is busy
at a particular point in time, then all of the parts being
processed will tend to have large times in system (i.e.,
the times are positively correlated).

Law and Kelton (2000, pp. 283-290) discuss inspec-
tion, confidence-interval, and time-series approaches that
might possibly be used for comparing model and system
output data.
3 GUIDELINES FOR OBTAINING GOOD
MODEL DATA
A model is only valid for a particular application if its
logic is correct and if it uses appropriate data. In this sec-
tion we provide some suggestions on how to obtain good
model data.
3.1 Two Basic Principles
If a system similar to the one of interest exists, then data
should be obtained from it for use in building the model.
These data may be available from historical records or may
have to be collected during a time study. Since the people
who provide the data might be different from the simula-
tion analysts, it is important that the following two princi-
ples be followed:

The analysts need to make sure that the data
requirements (type, format, amount, why needed,
etc.) are specified precisely to the people who
provide the data.
The analysts need to understand the process that
produced the data, rather than treating the obser-
vations as just abstract numbers. For example,
suppose that data are available on the time to load
a ship, but there are a few observations that are
significantly larger than the rest (called outliers).
Without a good understanding of the underlying
process, it is impossible to know whether these
large observations are the result of measuring or
recording errors, or are just legitimate values that
occur with small probability.
3.2 Common Difficulties
The following are four potential difficulties with data:

Data are not representative of what one really
wants to model

Example 10. The data that have been collected during
a military field test may not be representative of actual
combat conditions due to differences in troop behavior
and to lack of battlefield smoke.

Data are not of the appropriate type or format

Example 11. In modeling a manufacturing system, the
largest source of randomness is usually random down-
times of a machine. Ideally, we would like data on
time to failure (in terms of actual machine busy time)
and time to repair of a machine. Sometimes data are
available on machine breakdowns, but quite often they
are not in the proper format. For example, the times to
failure might be based on wall-clock time and include
periods that the machine was idle or off-shift.

Data may contain measuring, recording, or round-
ing errors

Example 12. Data representing the time to perform
some task are sometimes rounded to the closest 5 or
10 minutes. This may make it difficult to fit a con-
tinuous theoretical probability distribution to the data,
since the data are now discrete.

Data may be biased because of self-interest

Example 13. The maintenance department in an
automotive factory reported the reliability of certain
machines to be greater than reality to make themselves
look good.
4 SUMMARY
All simulation models need to be validated or any deci-
sions made with the model may be erroneous. The follow-
ing are the ideas that we believe are the most important for
developing a valid and credible model:

Formulating the problem precisely
Interviewing appropriate subject matter experts
Interacting with the decision-maker on a regular
basis throughout the simulation project to ensure
that the correct problem is being solved and to
promote model credibility
Developing a written conceptual model
28
Law and McComas

Performing a structured walk-through of the con-
ceptual model for a nonexistent system, this
may be the single most-important validation tech-
nique
Performing sensitivity analyses to determine im-
portant model factors
Comparing model and system results for an exist-
ing system (if any) this is, in general, the most
definitive validation technique available
Reviewing of model results and animations to see
if they appear to be reasonable

Many of the above ideas would seem to be just com-
mon sense. However, our experience indicates that they
are very often not applied.
REFERENCES
Balci, O. 1998. Verification, Validation and Testing, in
The Handbook of Simulation, J. Banks, ed., Chapter
10, John Wiley, New York.
Banks, J., J. S. Carson, B. L. Nelson, and D. M. Nicol.
2001. Discrete-Event System Simulation, Third Edi-
tion, Prentice-Hall, Upper Saddle River, N. J.
Carson J. S. 1986. Convincing Users of Models Validity
Is Challenging Aspect of Modelers Job, Industrial
Engineering, 18: 74-85.
Law, A. M. 1991. Simulation Study Puts the Right Heat
On at Kaiser Aluminum, Industrial Engineering, 23:
16-17.
Law, A. M. and W. D. Kelton. 2000. Simulation Modeling
and Analysis, Third Edition, McGraw-Hill, New York.
Montgomery, D.C. 2000. Design and Analysis of Experi-
ments, 5th Edition, John Wiley, New York.
Sargent, R. G. 1996. Verifying and Validating Simulation
Models, Proceedings of the 1996 Winter Simulation
Conference, San Diego, 55-64.
Shannon, R. E. 1975. Systems Simulation: The Art and Sci-
ence, Prentice-Hall, Englewood Cliffs, N.J.
AUTHOR BIOGRAPHIES
AVERILL M. LAW is President of Averill M. Law &
Associates, a company specializing in simulation consult-
ing, training, and software. He has been a simulation con-
sultant to numerous organizations including Accenture,
ARCO, Boeing, Compaq, Defense Modeling and Simula-
tion Office, Kimberly-Clark, M&M/Mars, 3M, U.S. Air
Force, and U.S. Army. He has presented more than 335
simulation short courses in 17 countries. He has written or
coauthored numerous papers and books on simulation, op-
erations research, statistics, and manufacturing including
the book Simulation Modeling and Analysis that is used by
more than 75,000 people. He developed the ExpertFit dis-
tribution-fitting software and also several videotapes on
simulation modeling. He has been the keynote speaker at
simulation conferences worldwide. He wrote a regular
column on simulation for Industrial Engineering maga-
zine. He has been a tenured faculty member at the Univer-
sity of Wisconsin-Madison and the University of Arizona.
He has a Ph.D. in industrial engineering and operations re-
search from the University of California at Berkeley. His
E-mail address is [email protected] and his
Web site is www.averill-law.com.
MICHAEL G. MCCOMAS is Vice President of Averill
M. Law & Associates for Consulting Services. He has
considerable simulation modeling experience in applica-
tion areas such as manufacturing, oil and gas distribution,
transportation, defense, and communications networks.
His educational background includes an M.S. in systems
and industrial engineering from the University of Arizona.
He is the coauthor of seven published papers on applica-
tions of simulation.
29