Ventana software support forum

Posted by Jim Duggan <JAMES.DUGGAN@NUIGALWAY.IE>
Hi,

At this years Boston conference, I noticed that a
number of presenters made use of existing and
well known SD models as part of their work (for
example, the long wave economic model, the
boom and bust model, the beer game...)

In other disciplines, for example computer science,
it is common to have an established ""library"" of test cases/models
from which new techniques can be benchmarked. For example, if someone
developed a new algorithm for doing x, they could run this against a
common test case/model and so demonstrate whether their
approach is an improvement on existing techniques.

So my question is: are there a core set of SD models
that could be used as benchmarks for researchers involved
in SD analysis research (for example, eigen value analysis,
optimisation, decision analysis,... etc.), and who need to benchmark the
efficiency and effectiveness of these techniques against existing approaches?

best regards,
Jim Duggan.

Dr. Jim Duggan
Lecturer, Department of Information Technology,
NUI, Galway.

http://corrib.it.nuigalway.ie
Posted by Jim Duggan <JAMES.DUGGAN@NUIGALWAY.IE>
posting date Mon, 08 Aug 2005 23:37:26 +0100

Posted by Joel Rahn <jrahn@sympatico.ca>
Two books that contain several standard or 'classic' models that are
often used as 'test-beds' and both are available from Pegasus:

Study Notes In System Dynamics by Michael R. Goodman
Managerial Applications of System Dynamics by Edward B. Roberts

Goodman has examples of small models, complete with model listings (in
DYNAMO, but easily translated to Vensim), and uses them as the basis for
a workbook of questions and answers about modeling and understanding
dynamics. Roberts has several models (with equations) from early work in
industrial dynamics including classics like ""Market Growth as Influenced
by Capital Investment"" and ""A Simple Model of Project Dynamics"".

Recent textbooks like those by John Sterman and Bob Cavana also contain
many examples of such 'standard' models.

My personal favorite for testing ideas is the ""Market Growth"" model-not
trivially small, not dauntingly complicated, can be divided into
meaningful sectors, has all the hallmarks of SD models: delays,
non-linearities, interesting behavior, real-world parameters...

Joel Rahn
Posted by Joel Rahn <jrahn@sympatico.ca>
posting date Tue, 09 Aug 2005 09:13:21 -0400

Posted by ""Christian Erik Kampmann"" <cek.marktg@cbs.dk>
Jim, I don't believe there is any formal selection of 'core' test models.
No doubt it is useful to have a collection of widely known and accepted
models, for several reasons:

1. It saves time and space explaining the purpose, structure and behavior of
the model
2. The 'core' models become so because they exhibit behavior that is
interesting or ubiquitous, hence increasing the a priori relevance of any
further analysis
3. A 'core' collection enhances the build-up of a shared knowledge base
beyond the general modeling principles and techniques that to a large extent
defines the field. It lies somewhere between the substantive knowledge of
most fields and the purely methodological ""knowledge"".
4. When various modifications and analytical techniques are applied to the
same population of models, it is easier to compare them.

That said, I would think one shouldn't be too formalistic about defining the
core collection. Otherwise, one runs the risk of narrow vision and
self-referential studies.

Those were my impulsive thoughts on the matter.

Christian Erik Kampmann
Associate Professor
Center for Applied Management Studies
Copenhagen Business School
Birk Centerpark 40
DK-7400 Herning
Denmark
Posted by ""Christian Erik Kampmann"" <cek.marktg@cbs.dk>
posting date Tue, 9 Aug 2005 15:52:04 +0200