Page 129 - Getting the Picture Modeling and Simulation in Secondary Computer Science Education

and between and among different types of data.”) (Sargent, 2013). Furthermore, a model can be compared to other models (Bungartz et al., 2014; Wilensky & Rand, 2015), and finally, its consistency with previously verified theories can be checked (Bungartz et al., 2014). Notably, when a real system does not exist, a model can still be considered valid while not accurate (Schmid, 2005).
When interpreting the model’s outcomes, the judgement about the model’s validity can be based on objective criteria that use statistical tests on the outcomes of a model (Law, 2009; Sargent, 2013), or on subjective criteria.
Students’ Understanding and Difficulties
Test the model
Generate outcomes
Observe outcomes
Interpret outcomes
Stress test
Parameter variability – sensitivity analysis
Extreme condition test Degenerate tests
Animation
Trace
Operational graphics Performance measures
Techniques to interpret data
Approaches to data interpretation
Comparison to other models Comparison to real data
Predictive validation
Event validity 6
Historical data validation
Checking input - output consistency
Data relationship correctness
Checking consistency with previously verified theories
Objectively
Subjectively
for example comparizon using statistical tests
Figure 10: Testing a model — various techniques
In addition to these techniques, the modelers can involve various stakeholders to support the validation process. A peer group can be involved to assess the model’s correctness through a structured walkthrough (Law, 2009; Sargent, 2013). The customer ordering the model can be asked to participate in the review of the model (Carson & John, 2004) or domain experts can be consulted to assess the face validity of a model (Sargent, 2013; Wilensky & Rand, 2015) through, for example, Delphi test involving a panel of experts (Carley, 1996) or Turing test where experts
127