 |
Professor
James Evans,
University of Cincinnati |
Bio: James Evans academic background is in Industrial Engineering,
having received BSIE and MSIE degrees from Purdue, and a Ph.D. from
Industrial and Systems Engineering at the Georgia Institute of Technology,
and has been with University of Cincinnati since 1975. Dr. Evans
has published many journal articles and is the author of a variety
of books in the areas of management science, quality management,
statistics and simulation, operations management, and creative thinking.
Dr. Evans consulting experience includes operations research applications
for AT&T, Kroger, Structural Dynamics Research Corporation,
The American Baseball League (I scheduled the umpires for over 20
years until the 2000 season), and Procter & Gamble; creative
thinking training for Hallmark Cards; technical writing for Decisioneering,
Inc.; and statistical process control training and quality management
consulting for several companies and organizations in Ohio, particularly
in understanding and applying Baldrige Criteria for Performance
Excellence and applying for the Baldrige Award.
Q: What classes are you teaching with Crystal
Ball (CB)?
Evans: We have a two course core sequence in quantitative
methods in our MBA program: Statistics for Managers (2 cr. hours)
and Decision Models and Optimization (4 cr. hours). I briefly introduce
CB in the statistics class when discussing Monte Carlo methods in
statistics, and also use it to illustrate sampling distributions.
In the Decision Models class, I spend a full week using in to support
a class on risk analysis, and also use it to replicate simple system
simulation models of queuing and inventory created
in Excel. In the future, I plan on integrating it more extensively
in the statistics class to provide more experiential insights into
the theoretical concepts.
> Download a sample PowerPoint
presentation from one of Dr. Evan's classes 
Q: How are you teaching the software in your
classes?
Evans: One of the things we've done recently in our MBA
program is to increase the number of credit hours relative to class
contact hours. This was done by design to "off-load" some
topics to alternative modes of learning, such as the Web or off-line
tutorials and shorten the program cycle time.
I expect students to work through the Crystal Ball tutorial outside
of class to learn the mechanics of using CB (using Decisioneering's
tutorial), and then do only a brief review in class in order to
focus more on simulation concepts and applications. Because CB is
so easy to learn and use, I haven't found any students who have
had much difficulty in mastering the mechanics quickly. (Creating
the correct models, on the other hand, is another story!) I also
use half the class for a hands-on lab session to allow them to identify
any questions they might have and apply the software to a problem.
Q: How does Crystal Ball help your students?
Evans: Let me back up a few years. I came across Crystal
Ball soon after it was first released. Simulation has been a part
of our quantitative methods class as long as I can remember. With
my engineering degree background, my focus was primarily on system
simulation principles - process flows, and so on - using SLAM or
a similar language.
Crystal Ball opened up an entire new scope of applications to business
students, particularly in the financial arena. Because we use Excel
extensively in all our courses in the college, Crystal Ball was
a natural add-in that made the quantitative course much more relevant
to a business degree, and tied in closely with many other courses
in their program. I know that students have used it in subsequent
finance, accounting, and operations management courses. So in short,
the answer is that CB makes their learning about simulation very
relevant to their program.
In addition, many of our MBA students are part-timers with full-time
jobs. I'd say that a very large proportion of them have told me
that they have seen many potential applications for CB in their
work, and many have used it beneficially in their jobs. One of our
college's strategies is "experiential learning." As such,
I have a course requirement of a project using a real or hypothetical
situation from their work experience. Approximately half the students
do something using Crystal Ball, and many of the projects have been
quite creative. It clearly has helped them apply many of the principles
they learn in their classes to real situations.
Q: Why do you like the functionality of Crystal
Ball?
Evans: The interface is highly intuitive and the software
is extremely easy to use. I especially like the function extensions
for random variate generation and extracting CB output data into
customized Excel spreadsheets.
Q: Are you using any
publications that reference Crystal Ball?
Evans: Of course! I use a very good book by two guys named
Evans and Olson...Statistics,
Data Analysis and Decision Modeling, 2E! Seriously, I've
authored or coauthored several textbooks in management science and
simulation (Evans and Olson, Introduction
to Simulation and Risk Analysis, 2E, Prentice-Hall), all
of which have used Crystal Ball. I've been quite pleased with the
support I've received from Decisioneering as both an author and
instructor.
Q: Do
you have any example models that you may have that you would be
willing to share with readers?
Sure. A simple Monte Carlo experiment that arises in quality control
is explained in my book, Statistics, Data Analysis, and Decision
Modeling, 2nd Edition (with D. Olson), Prentice-Hall, 2003.
In using control charts, statisticians have developed factors to
estimate the standard deviation from the range using a factor called
d2 (note this is d sub 2). By dividing the range by d2 (which depends
on the sample size), we obtain an estimate of the standard deviation.
It's usually a mystery where these factors come from, but a simple
Monte Carlo experiment can easily illustrate it. In the model we
assume a mean of 0 and n = 5. Each value in the range B4:B8 is computed
using the function =CB.NORMAL(0,$B$1). Cell B10 is defined as a
Crystal Ball forecast. For each trial, new values of the five samples
are generated, and the value of the forecast cell is recorded.
The CB forecast chart shows the distribution of R/s. For one simulation
of 10,000 trials, the sample mean, that is, the estimate of d2 turned
out to be 2.34. Published statistical tables give this value as
2.326, so we see that the simulated value is very close to the value
identified by statisticians. We also see that the actual distribution
of R/s covers a fairly wide range, indicating that there is considerable
potential for error in the estimate. This is usually not discussed
in the quality control literature, but CB makes it easy to illustrate
this point.
> Download the Quality Control
Model
I also use a more entertaining example of roulette (also attached)
when I begin to lecture on risk analysis.
> Download the Roulette Model
|
