Database Reference
In-Depth Information
the sensitivity of model results to different model specifications. Later chapters will
take up the issue of sensitivity analysis in more detail.
If enough data are available, check your model's ability to use known initial
data and “predict” known recent data for the same variables. This calibration and
verification step lets you project scenarios with greater certainty.
1.9 The Detailed Modeling Process
Following is a set of easy-to-follow but detailed modeling steps. These steps are not
sacred: they are intended as a guide to get you started in the process. You will find
it useful to come back to this list once in a while as you proceed in your modeling
efforts.
1. Define the problem and the goals of the model. Frame the questions you want
to answer with the model. If the problem is a large one, define subsystems of
it and goals for the modeling of these subsystems. Ask yourself: Is my model
intended to be explanatory or predictive?
2. Designate the state variables. (These variables will indicate the status of the
system.) Keep it simple. Purposely avoid complexity in the beginning. Note the
units of the state variables.
3. Select the control variables—the flow controls into and out of the state variables.
(The control variables are calculated from the state variable in order to update
them at the end of each time step.) Note to yourself which state variables are
donors and which are recipients with regard to each of the control variables.
Also, note the units of the control variables. Keep it simple at the start. Try to
capture only the essential features. Put in one type of control as a representative
of a class of similar controls. Add the others in step 10.
4. Select the parameters for the control variables. Note the units of these parame-
ters and control variables. Ask yourself: Of what are these controls and their
parameters a function?
5. Examine the resulting model for possible violations of physical, economic, and
other laws (for example, any continuity requirements or the conservation of
mass, energy, and momentum). Also, check for consistency of units. Look for
the possibilities of division by zero, negative volumes or prices, and so forth.
Use conditional statements if necessary to avoid these violations.
6. To see how the model is going to work, choose some time horizon over which
you intend to examine the dynamic behavior of the model, the length of each
time interval for which state variables are being updated, and the numerical
computation procedure by which flows are calculated. (For example, choose in
the STELLA program Time Step
24.) Set up a graph and
guess the variation of the state variable curves before running the model.
7. Run the model. See if the graph of these variables passes a “sanity test.” Are
the results reasonable? Do they correspond to known data? Choose alternative
=
1, time length
=
