UDK 005.

4:004
Original scientific paper

'(9(/23,1*6<67(0'<1$0,&602'(/6:,7+67(3-BY-67(3
APPROACH
MLUMDQD3HML-%DFK9ODWNRHUL
University of Zagreb, Faculty of economics, Zagreb, Croatia
{mpejic,vceric}@efzg.hr

Abstract: System dynamics is a powerful tool that enhances learning about company,
market and competitors; portrays the cognitive limitations on the information gathering and
processing power of human mind; facilitates the practice of considering opinions; and
supports building of "What if" scenarios. Although the literature on system dynamics
modeling is very rich with applications in many fields, not many papers on developing
system dynamics models were published so far. In this paper we portray current approaches
to the development of system dynamics models. These are (1) model development based on
influence diagram, (2) model development based on the identification of resources and their
states, (3) usage of generic structures for specific domain field, and (4) component strategy
for the formulation of system dynamics models. Validation is an important issue that none of
WKHVHDSSURDFKHVWDFNOHV:HSURSRVHDVWHS-by-VWHSDSSURDFKWKDWLQWHJUDWHVYDOLGDWLRQ
with developing process of system dynamics models. This approach will be demonstrated on
the example of development of a simple inventory model.

Keywords: system dynamics, model development, validation, inventory.

1. INTRODUCTION
The process of system dynamics model development is not simple, and a not many
papers about this topic were published so far. Those who are just beginning to deal with
system dynamics can easily be mislead by the simplicity of the system dynamics
development software, and may attempt to develop the model in one step. However, this type
of approach often results with the model containing various faults which are difficult to
correct.
System dynamics models can help in understanding structure and behavior of the system
with nonlinear links and feedback. However, experience in development of system dynamics
model teach us that proper understanding of the model behavior is very hard to achieve if the
DSSURDFKWRWKHGHYHORSPHQWRIWKHPRGHOLVQRWJUDGXDO7KHUHIRUHZHUHFRPPHQGVWHS-byVWHS DSSURDFK WR V\stem dynamics model development which integrates the evaluation of
the model with the process of model development. Such approach enables better
understanding of the model behavior, as well as establishing better confidence in the model.
The goal of this SDSHULVWRGHVFULEHWKHVWHS-by-VWHSDSSURDFKWRGHYHORSPHQWRIWKH
system dynamics model, and demonstrate it on the example of the inventory model. The
paper consists of the following parts. After the introduction, in the second part of the paper
171

M. 3HML-%DFK9HULDeveloping system dynamics models ...

current approaches of system dynamics model development are shown. In the third part
VWHS-by-VWHS DSSURDFK RI system dynamics model development is described, while the
fourth part presents development of the inventory model using this approach. The last part of
the paper gives conclusion.
2.

DEVELOPMENT AND VALIDATION OF SYSTEM DYNAMICS

MODELS

Current approaches to the development of system dynamics models are: (1) model
development based on casual-loop diagram (Coyle, 1996), (2) model development based on
the identification of resources and their states (Wolstenholme, 1990), (3) usage of generic
structures for specific domain field (Wolstenholme, 2004), and (4) component strategy for
the formulation of system dynamics models (Forrester, 1968; Goodman, 1975). Validation is
an important issue that none of these approaches tackles.
Model development based on influence diagrams proposes building quantitative model
with system dynamics software using causal-loop diagrams. Casual-loop diagrams are very
suitable for explaining model structure to management at the beginning and at the end of the
modeling process. However, some problems may arise in causal-loop diagramming, both in
development of causal-loop diagrams and in the deriving system behavior from them. The
main problem is that causal-loop diagrams obscure the stock and flow structure of systems
(Richardson, 1986). Casual-loop diagrams are then used for deriving of both stock and flow
diagrams, as well as system dynamics equations.
System dynamics approach is based on identification of resources, their states and rates
at which resources change their states. Resources (levels or stocks) could be material, people,
cash, orders, etc. A state of the resource can be defined as any accumulation of the resource
which is relevant to the purpose of the model. The rate at which resources are converted
between states is represented by rate variables. Wolstenholme (1990) proposes creating the
structure of systems with the goal of recognizing resources and states. He proposes
identification of relevant resources related to the modeling goal, as well as states and rates at
which resources change. Based on this, stock and flow diagram and model equations are
derived.
Generic structures are relatively simple structures that occur in various situations (Albin
et al., 2001), and can help with the creation of dynamic hypotheses at the front end of the
modeling process as well as with communication on systemic insights at the back end of the
modeling process. In practice, it is often beneficial to use the archetypes in parallel
throughout the process to guide high-level thinking whilst detailed modeling is taking place
(Wolstenholme, 2004). In that way, stock and flow diagrams are created without any
preliminary preparation. However, by simply fitting the system to a generic structure, the
inexperienced modeler can easily use wrong generic structures that are not suitable for
particular system. (Breirova, 2001).
The most recent concept is the component strategy to the development of system
dynamics models. This approach concentrates on the formulation of the Forrester stock and
flow diagram, and incorporates the concept of an interaction matrix to assist in formulation
of such models (Burns, et al., 2002). In this strategy the quantities that will be included in the
model and their associated interactions are generated simultaneously. The goal of
introducing this strategy was to develop computer aids that could facilitate model
formulation in order to speed up the process of system dynamic model formulation. This
strategy would then divide the labor of modeling into human and computer part, where
computer part could be automated.

172

Journal of information and organizational sciences, Volume 31, Number 1 (2007)

System dynamics models are used in analyzing the structure and the behavior of the
system as well as for designing efficient policies of managing the system. For example, using
system dynamics model of decisions can help in finding appropriate decisions for the
company (Merten, 1991; Morecroft, 1984). Moreover, these models have a significant role in
the education of managers (Graham et. al, 1992). Clients and other potential users obviously
want to be sure that they can trust the system dynamics model, because model with
significant flaws can lead them to wrong decisions (Richardson, 1996).
Tests for acquiring confidence in the system model dynamics can be divided into two
groups: (1) structure tests and (2) behavior tests (Forrester, et.al, 1979). Structure tests
(structure verification test, parameters verification test, extreme conditions test, model border
adequacy test and dimensional consistency test) compare the structure of the system
dynamics model with the structure of the real system so that every relationship between the
elements of the real system is being compared with the relationship between corresponding
elements of the model which is described by mathematical equation. Behavior tests (behavior
reproduction test, behavior prognosis test, behavior anomaly test, generic behavior test,
extreme policy test, border adequacy test and behavior sensibility test) are conducted to
determine whether the behavior of the model matches the behavior of the real system, and
here the relationship between the structure and the model behavior is analyzed with
particular care.
Interviewing was introduced as another strategy for assessing of system dynamics
models (Diker et al, 2005), based on importance of using expert judgment for assessing
purposes. The paper presents four illustrations about the use of interviews in the validation of
system dynamics models. These four methods differ in a number of points: who was
interviewed, technology of delivery, type of questions, how behavior was presented, how
structure was presented, data processing and data analysis. These methods provide a number
of question formats and analysis techniques that could be used in the validation process.

3. 67(3-BY-67(3$33ROACH
Evaluation is a process in which users acquire confidence in the system dynamics model
(Richardson et.al, 1981). The experience shows that it is very important that the process of
model evaluation is conducted in parallel to the development of the model, rather than after
the model completion. It means that evaluation of the model should be an iterative procedure
conducted during all phases of the simulation modeling. This is especially important since it
is well known that too fast model development is WKH FRPPRQ EHJLQQHUV PLVWDNH. Most
frequently beginners develop whole models in a single stage, and conduct evaluation tests
only when the model is already finished. This approach cannot guarantee development of
high quality and robust model whose behavior and structure matches reality. Because of the
formerly mentioned problem with the use of casual-loop diagrams (causal-loop diagrams
obscure the stock and flow structure of systems) we recommend the development of the
stock and flow diagram right after the system analysis.
Because of all these we recommend development of the system dynamics model in
several steps:
1.

Development of the basic model

2.

Conducting the basic evaluation tests extreme condition tests, behavior sensibility
test and dimension consistency test

3.

Expansion of the model with one or more feedbacks

173

M. 3HML-%DFK9HULDeveloping system dynamics models ...

4.

Re-conducting aforementioned evaluation tests for the new version of the model

5.

If (a) these tests are not giving satisfactory results or if (b) the user on the basis of
understanding the system reach the conclusion that it is necessary to expand the
model with new feedbacks, step two is repeated and the whole procedure is
continued

6.

If the results of the aforementioned tests are satisfying, and the modeler concludes
that the model is complete, the other evaluation tests mentioned before are carried
out

Therefore VWHS-by-VWHSDSSURDFKWRV\VWHPG\QDPLFVPRGHOdevelopment proposes the

use of the three basic evaluation tests which point out to the modeler the existence of errors
and oversights (dimensional consistency test and extreme conditions test), and also help in
understanding the influence of every variable on the model behavior (behavior sensibility
test). Short description of these tests follows.
Dimensional consistency test
In the system dynamics model it is important that the units of measure of variables on
both sides of the equation are equal. This test also checks whether dimensions of variables in
the model correspond to the unit in which they can meaningfully express the real variables
which exist in the company. The test is conducted using built-in function of program
language used for system dynamics model development.
Extreme conditions test
This test checks whether the structure of the model is such that the behavior of the
model in extreme conditions matches the behavior of the real system in same situations. For
example, if the demand for the company products is equal to zero during the whole
simulation, then the number of delivered product should also be zero, and there should not be
any revenues from the product sales as well as no directs costs related with the sales.
Behavior sensibility test
This test is focused on detecting the parameters whose small changes cause significant
change in the model behavior. The fewer such parameters, the higher the credibility of the
model is. However, the model behavior sensibility is acceptable if in the real system small
change of the parameter values also causes significant change of the system behavior. The
goal of the system dynamics is to find the parameters which have most effect on system
behavior, and can thus be used for.system management policies. If this test shows that the
model is not sensible to the changes of some parameters, it can be concluded that for
assessment of these parameters subjective judgment is reliable enough.

174

Journal of information and organizational sciences, Volume 31, Number 1 (2007)

4. (;$03/('(9(/23,1*6,03/(,19(1725<02'(/%<67(3BY-67(3$3352$&+
We describe here development of the inventory model in the company which solely
imports products, and GRHVQW KDYH any production capability. Model was developed using
Vensim system dynamics software.
The development of the inventory model is carried out through three steps. In the first
step the simple model of the inventory with supply and delivery is analyzed. ThiV PRGHOV
disadvantage is that in it the inventory can become negative. In the second step feedback is
added, which prevents the inventory to become negative. In the third step ordering new
products is added into the model.

4.1. SIMPLE INVENTORY MODEL

Inventory model in the beginning of the modeling process consists only of the inventory
level and the speed of supply and delivery (Figure 1). Supply equals 1000 product per month,
the same as the monthly delivery of the products. In the beginning of the simulation there are
1000 products in the inventory, and the model is balanced, i.e. during the whole simulation
inventory contains 1000 products.

inventory
supply

delivery

Figure 1. Simple inventory model flow diagram

The equations of the model are:

Evaluation of the model is done in the following way. Dimensional consistency test is
conducted by the 9HQVLPV built-in function, and it shows that the level units and the speed
are dimensional consistent. Extreme conditions test is carried out with two assumptions: (1)
delivery=0 and (2) supply=0.
It was shown that when delivery=0, inventory grows linearly (Figure 2). This kind of
behavior is consistent with the situation when company does not succeed in selling the
products, but still keeps on buying new ones. Since this kind of behavior is not realistic, the
model needs to be expanded so that the process of buying new products is formulated on the
basis of the demand information.
175

M. 3HML-%DFK9HULDeveloping system dynamics models ...

Graph for inventory
60,000
1
1

45,000

1
1
1
1

30,000

1
1
1

15,000

1
1
1

1
1

16

Inventory: delivery = 0

24
32
Time (Month)

40

48

Figure 2. Model behavior when delivery =0

If the company does not succeed in supplying new products (i.e. supply is equal to zero),
the inventory will be depleted and number of products decreases until it falls to zero.
However, in the model inventory keeps on decreasing and even becomes negative (Figure 3),
and this is unreal. Because of that the model needs to be corrected so that the delivery of the
inventory is limited, and this is done in the next step.
Graph for Inventory
1,000

1
1
1
1
1

-14,250

1
1
1

-29,500

1
1
1
1

-44,750

1
1

-60,000
0

Inventory : supply = 0

16

24
32
Time (Month)
1

40

48

Figure 3. Model behavior when supply=0

4.2. LIMITING INVENTORY DELIVERY

In inventory modeling we should take in consideration that the company will not always
be able to satisfy the demand for its products, and that it can sell only the amount of products
that it has in warehouse. Until the number of products in inventory is higher than the desired
number of products in inventory, sale equals demand. When the inventory decreases,
management restricts delivery. However, it does not deliver products to the first customers
that appear, but always keeps a few products for its permanent buyers. If the demand is
constantly higher than the supply, the inventory will gradually decrease until the effect of the
inventory state doesnt stop further delivery (Figure 4).

176

Journal of information and organizational sciences, Volume 31, Number 1 (2007)

DESIRED INVENTORY

inventory ratio

effect of
inventory
ratio

effect of inventory ratio

lookup

Inventory
supply

delivery

DEMAND

Figure 4. Flow diagram of the inventory model with delivery constrains

The model is expanded with following equations:

The function of the inventory status is presented as follows (Figure 5). The abscissa
shows a current/desired inventory ratio, while the ordinate shows the effect of the inventory
state. The number of delivered products is calculated as a product of inventory state effect
and demand. The effect of inventory state depends on the ratio of current/desired number of
products on stock. For example, if the current/desired number of products ratio is 3/10, the
effect of inventory state is 0.5, what means that the management delivers half of the desired
quantity. The smaller current/desired inventory radio, management delivers a smaller part of
the requested quantity. If there inventory is empty, current/desired inventory radio equals to
zero. In this case the effect of inventory status also equals to zero, and management does not
delivery any products.

177

M. 3HML-%DFK9HULDeveloping system dynamics models ...

Figure 5. Effect of inventory ratio lookup

If both the demand for the products and its supply is equal to 1000 products per month
with, the model would be balanced during the entire simulation. Now let us assume that the
demand is 1100 products monthly instead of 1000 products. At he beginning of the
simulation management has 1000 products in inventory, i.e. equal to the desired inventory.
Although demand is 1100 products, management keeps on supplying only 1000 product per
month. Since the delivery is larger than supply, inventory is gradually decreasing. Assume
WKDW WKH FRPSDQ\V SROLF\ LV to delivery requested number of products until inventory
decreases to approximately 1/3 of desired inventory, and after that management reduces
delivery until the number of 1000 products monthly is attained and inventory reaches
equilibrium value (Figure 6).
1,000 P
1,200 P/Month
1,200 P/Month

3
3

500 P
1,000 P/Month
1,000 P/Month
0
800

P
P/Month
P/Month

16

24
32
Time (Month)

40

48

1
1
1
1
1
1
1
1
1
1
Inventory : Current1
P
2
2
2
2
2
2
2
2
2
supply : Current
P/Month
3
3
3
3
3
3
3
3
3 P/Month
delivery : Current

Figure 6. Inventory model behavior with the demand of 1100 product per month
As in the previous step of model development dimensional consistency test is conducted
with the Venism software built-in function, and it is shown that the level units and the speed
are dimensionally consistent.
Extreme conditions test is conducted again, with two assumptions: (1) supply = 0 and
(2) demand = 0.
If the management stops buying the products and all other parameters are left unchanged
(inventory and desired inventory are 1000, demand is 1000) the inventory will decrease until
all he products are sold, i.e. until they reach the value 0, as seen in Figure 7.

178

Journal of information and organizational sciences, Volume 31, Number 1 (2007)

Graph for Inventory
1,000

750
1

500
250

Inventory: supply=01

3
Time (Month)
1

Figure 7. Model behavior when supply=0

If the demand is 0 and the management keeps on supplying 1000 products every month,
the inventory should linearly increase, which is visible from the graph of Figure 8.
Graph for Inventory
60,000
1
1

45,000

1
1
1
1

30,000

1
1
1

15,000

1
1
1

1
1

Inventory : Current 1

16

24
32
Time (Month)
1

40

48

Figure 8. Model behavior when demand = 0

In order to conduct the VHQVLWLYLW\WHVWWKHYDOXHRIIXQFWLRQRILQYHQWRU\VWDWXVHIIHFWRQ
GHOLYHU\ ZDV FKDQJHG (Figure 9). Current form of the function reflects the position of
management on backup inventory. However, there are no exact rules in the company
concerning backup inventory for permanent buyers. So the function is changed in order to
reflect both liberal and restrictive policy of keeping the backup inventory. If the nonlinear
function is shifted to the right, it reflects the restrictive policy of keeping the backup
inventory, because delivery is starting to decrease with the larger value of inventory ratio. In
this case the management will keep higher level of inventory if the delivery increases. The
reverse is true if the nonlinear function is shifted to the left.

179

M. 3HML-%DFK9HULDeveloping system dynamics models ...

Figure 9. Function for the basis of sensitivity test

The simulation is carried out with changed function of inventory status effect on
delivery, which reflects current, liberal and restrictive policy of keeping backup inventory. It
is assumed that the demand for the products is 1100 products during the whole simulation.
The model behaves according to expectations. When the policy of inventory delivery is more
liberal than the current one, the equilibrium value of inventory is lower. The reverse is true if
the policy of inventory delivery is more restrictive (Figure 10).
Graph for Inventory
1,000

1
2

750

3
1

500

23

250

3
1

1
2

0
0

1
2

1
2

16

1
2

1
2

1
2

1
2

24
32
Time (Month)

1
2

1
2

1
2

40

1
2

1
2

48

Inventory : Current 1
1
1
1
1
1
1
1
1
1
1
Inventory : liberal policy 2
2
2
2
2
2
2
2
2
Inventory : restrictive policy
3
3
3
3
3
3
3
3

P
P
P

Figure 10. Model behavior with sensitivity test

4.3. ORDERING PROCESS

In previous step it was assumed that management always orders the same amount of
products regardless of the change in the demand. This kind of assumption is not realistic, and
therefore the model will be expanded in order to represent the process of ordering in the
company.
Suppose that cRPSDQ\V PDQDJHPHQW orders every month the amount of products that
was sold in the previous month in order to fill the inventory. The supplier needs
approximately 6 weeks to delivery the ordered goods. Management takes into account the
desired inventory, compares current inventory with desired inventory and orders product
every month to eliminate the difference between them. The desired inventory depends on the
demand for the products, and management wants to keep the quantity of products that is
enough to settle the demand during 6 weeks. The ordering process is developed (Figure 11)
according the Guided Study Program in System Dynamics (1999).

180

Journal of information and organizational sciences, Volume 31, Number 1 (2007)

The model contains two negative feedbacks:
1) Increasing the inventory causes the increase of inventory ratio. The higher the inventory
ratio, the larger the effect of inventory status on delivery, and delivery grows. Because
of increased delivery, the inventory decreases.
2) Increasing the inventory decreases inventory deviation. The smaller the inventory
deviation, management orders less new products. Because of the smaller supply,
inventory decreases.
DESIRED INVENTORY COVERAGE

<expected demand>

desired inventory

inventory gap
effect of inventory ratio
inventory
effect of inventory
ratio lookup
Inventor
supply

delivery

products ordered
products delivered

DEMAND

DELIVERY TIME
TIME TO CORRECT
INVENTORY GAP
replacement ordering
new orders

Figure 11. Flow diagram of inventory model with constraint of delivery and ordering
Evidently no company works under ideal condition in which demand is always constant
and inventory always equals desired inventory. Therefore we will test the behavior of the
model in conditions where demand increases only once. It is assumed that demand increases
after 10 months from 1,000 products monthly to 1,500 products monthly, and it remains
unchanged until the end of the simulation. It this case the demand equation is

After the demand increases from 1,000 products to 1,500 products monthly, desired
inventory also changes from 1,500 products to 2,250 products (Figure 12). Since for the first
10 months inventory equals 1,500 products, the management has to order new products so
the inventory could grow to the desired level. However, when management orders product to
remove the deviation of inventory it does not take in consideration the time necessary for
delivering products from the supplier, what causes system oscillations. After the demand
181

M. 3HML-%DFK9HULDeveloping system dynamics models ...

increases, the sales also grow and the inventory diminishes. Management compares current
inventory with desired inventory, and orders products to eliminate the difference. Upon next
order, management again compares current with desired inventory. However, the problem is
in that the products ordered in previous month have not still arrived. Because of that
management orders too many extra products. After the ordered products finally begin to
arrive, management realizes that the inventory grows too much so they order fewer products.
+RZHYHU DJDLQ WKH\ GLGQW take in account the products which are still on their way to
inventory, and they cut back on orders too much and this results in a large decrease of
inventory. So, because of the delay in delivery of products, management at first orders too
much, and later not enough products. However, inventory oscillations become smaller and
smaller, and inventory reaches new equilibrium level of 2,250 products.
Graph for Inventory
4,000

3,000
1

2,000

1
1

1,000

0
1

Invnetorye : Current 1

19

37
Time (Month)
1

54

72

Figure 12. Model behavior with the increase in demand

The dimensional consistency test, extreme conditions test and the sensibility test are
conducted again. Because of the limited amount of space here, only one sensibility test
results; the extreme conditions test, will be shown.
The model sensibility test is conducted under the assumption that the DEMAND equals 1000
products during first 10 month, after which it grows to 1500 products. The values of the next
parameters are changed:
TIME OF CORRECT INVENTORY GAP (Ordering time) = 1, 3 and 5 months
Initial value of inventory = 1000, 1500, 3000 products
DESIRED INVENTORY COVERAGE = 1, 3 and 6 months
DELIVERY TIME = 0.5, 1.5 and 3 months
Inventory is an important issue for every company. Large inventory represent cost and
GHFUHDVHOLTXLGLW\DQGVPDOOLQYHQWRU\FDQFDXVHWKHORVVRIFRPSDQ\VPDUNHW Therefore,
inventory should oscillate as less as possible. This is why the goal of above sensitivity tests is
to find out optimum combination of parameters that would ensure lowest oscillations.
Ordering time
Ordering times represent the speed of reaction on deviation of current from desired
inventory. The inventory model shows that management can influence only the value of time
182

Journal of information and organizational sciences, Volume 31, Number 1 (2007)

of ordering stocks, while initial value of inventory, desired coverage of inventory, and the
time of delivery depend on many external factors. They could intuitively believe that it is
better to react faster in ordering inventory, and that this would bring larger stability of
inventory. However, the sensitivity test shows that inventory oscillations are higher with the
shorter time of ordering, and vice versa (Figure 13).
Graph for Inventory
4,000
3,000
1 2

2,000
1 2 3 1 2 3

1,000

1
3

2 3 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3 1

0
1

19

37
Time (Month)

Inventory : Ordering = 1 month

Inventory : Ordering = 3 months
Inventory : Ordering = 5 months

1
2

1
2

54

1
2

1
2

1
2

72

1
2

1
2

P
P
P

Figure 13. Behavior of inventory with the ordering time of 1,3 and 5 months
Initial value of Inventory
Graph for Inventory
4,000
3,000

2,000

1,000

0
1

19

Inventory : Inventory = 1000

Inventory : Inventory = 1500
Inventory : Inventory = 3000

37
Time (Mjesec)

54

72
JP
JP
JP

Figure 14. Behavior of inventory with the Initial value of inventory of 1000, 1500, 3000
products

183

M. 3HML-%DFK9HULDeveloping system dynamics models ...

Desired Inventory Coverage
Graph for Inventory
20,000
15,000

10,000

5,000

0
1

19

37
Time (Mjesec)

54

Inventory : Desired coverage = 1 month

Inventory : Desired coverage = 3 months
Inventory : Desired coverage = 6 months

72
JP
JP
JP

Figure 15. Behavior of inventory with desired inventory coverage of 1, 3 and 6 months
Delivery time

Graph for Inventory

6,000
4,500

3,000

1,500

0
1

19

Inventory : Delivery = 0.5 months

Inventory : Delivery = 1.5 months
Inventory : Delivery = 3 months

37
Time (Mjesec)

54

72
JP
JP
JP

Figure 15. Behavior of inventory with delivery time of 0.5, 1.5 and 3 months

5. CONCLUSION
In the paper VWHS-by-VWHSDSSURDFKWRGHYHORSLQJV\VWHPG\QDPLFVPRGHOZDV
shown. This approach consists of the following steps. In the first step the initial
version of the model is designed, which is tested by basic evaluation tests:
dimensional consistency test, extreme condition tests and behavior sensibility test. In
the second step the model is expanded with feedbacks, and the expanded version of
the model is tested using the aforementioned tests. The second step is repeated until
the model functions satisfactory. After that the other standard structure and behavior
184

Journal of information and organizational sciences, Volume 31, Number 1 (2007)

tests are applied. This approach helps in achieving significant degree of confidence
and understanding of model EHKDYLRU7KHVWHS-by-VWHSDSSURDFKLVVKRZQRQWKH
example of development of the simple inventory model.
REFERENCES
[1] Albin, S. (1996). Generic Structures: First-Order Linear Negative Feedback. Available
at: http://sysdyn.clexchange.org/sdep/Roadmaps/RM4/D-4474-2.pdf
[2] Breirova, L. (2001). Mistakes and Misunderstandings: Use of Generic Structures and the
Reality of Stocks and Flows. Available at:
http://sysdyn.clexchange.org/sdep/Roadmaps/RM8/D-4646-2.pdf
[3] Burns, J.R., Ulgen, O. (2002). Component Strategy for the Formulation of System
Dynamics Models. System Dynamics Conference 2002. Available at:
http://www.systemdynamics.org/conf2002/papers/Burns2.pdf
[4] Coyle R.G. (1996). System Dynamics Modelling: A Practical Approach. Chapman &
Hall London.
[5] Diker, V., Luna-Reyes, L. F., & Andersen, D., L. (2005). Interviewing as a strategy for
the assessment of system dynamics models. 23rd International Conference of the System
Dynamics Society, Boston, MA
[6] Forrester, J.W. (1968). Principles of Systems. Wright-Allen Press, Massachusetts.
[7] Goodman, M..R. (1974). Study Notes in System Dynamics (Cambridge : Wright-Allen
Press).
[8] Guided Study Program in System Dynamics (1999). Assignment #25. Available at:
http://sysdyn.clexchange.org/gsp98/assignments/Gsp-a25.pdf
[9] Richardson, G.P. (1986). Problems with Casual-loop Diagrams. System Dynamics
Review, Vol. 2, No.2, 158-170.
[10] Wolstenholme, E.F. (1990). System Enquiry: A System Dynamics Approach. Wiley,
Chichester.
[11] Wolstenholme, E.F. (2004). Using generic system archetypes to support thinking and
modelling. System Dynamics Review Vol. 20, No. 4, (Winter 2004): 341356.
Received: 16 March 2007
Accepted: 26 October 2007

185