Page 128 - Getting the Picture Modeling and Simulation in Secondary Computer Science Education

Chapter 6
On the other hand, numerous validation techniques rely on testing to judge the validity of a constructed model depending on its behavior and outcomes that are generated as the input parameters are varied, and then observed and interpreted.
Generate outcomes
Test the model Observe outcomes
Interpret outcomes
Figure 9: Testing a model
To generate the outcomes, several tests can be performed. Examples of such test are stress test (i.e testing with wide range of parameters and random numbers) (Carson & John, 2004), parameter variability–sensitivity analysis (i.e. varying the values of parameters and observing the resulting outcomes to determine whether the relations correspond to those of the real system), extreme condition test (i.e. checking that the model’s structure and outputs are plausible for extreme and unlikely combinations of the system’s levels of factors) and degenerate tests (i.e. looking into the degeneracy of the model’s behavior) (Sargent, 2013).
To observe the generated outcomes, several methods can be employed. The generated outcomes can be observed visually through animation (i.e. graphical display of the model’s behavior as it is run) and trace (i.e observation of the behavior of a specific entity during the run of the model) (Sargent, 2013). Another way to observe the generated outcomes is by making use of quantitative measures expressed through operational graphics (i.e. graphical display of various performance measures as the model runs) (Sargent, 2013) or numerically through various performance measures (Law, 2009).
Finally, the outcomes are interpreted. Several techniques can be used to interpret the model’s outcomes. Examples of such techniques are comparison with real data through historical data validation or predictive validation and checking for event validity (i.e. comparing the ‘events’ occurring during the model run with those occurring in the real system), checking input-output consistency and data relationship correctness (i.e. making sure “Data relationship correctness requires data to have the proper values regarding relationships that occur within a type of data,
126