��������������������������
The final design step is to select the modeling technique . The representation language or broad type of technique , such as diagram or computer simulation , impacts on functionality because of differences in accessibility and how the model can be used . Different techniques within these broad groupings can be used in similar ways and therefore provide similar functionality . However , there are important implications for accuracy , as the assumptions of the technique must be appropriately matched with the characteristics of the target system ( discussed in detail below ).
���������������������
The Build phase converts the model design to the implemented model
( subject to revision ). Different representation languages have different steps during this phase . In particular , models using mathematical equations or computer simulation quantify relationships and require data to set the parameters . Even games can require data to make the rules sufficiently realistic to stimulate behavior of interest . In contrast , diagrams do not require the same type of specification and the Build phase may be relatively minor . Despite these differences , the objective of the build phase is the same for all representation languages : to add detail to the model design and clarify any elements where there is ambiguity or inconsistency . For diagrams and simple board games , the Build phase provides a fresh look , perhaps clarifying some language or connections , but primarily confirming that the model captures all the key issues identified during the design discussions and envisaged in the model purpose . This involves tracing through all the relationships to confirm that no key influences or constraints are missing , and that the connections make logical sense . This process should be undertaken jointly by subject matter experts and the modeler , with the latter focused on asking clarifying questions .
For other types of models , building the model is a substantial technical process where specialists implement the model design using the agreed modeling technique ( s ) and is primarily the responsibility of the modeler . However , it is important for the policy analyst to remain involved , as there are many decisions to be made in interpreting or revising design elements and adding detail . Changes are common , as the rigor of building the model exposes problems with the design . For example , analysis may show that the relationship posited between two variables does not exist , or data may turn out to be missing , requiring replacement of a variable with one that is related and for which there are data .
There are three general steps in the Build phase for a mathematical or simulation model :
• specifying the relationships as the model structure ;
• parameterizing or calibrating the relationships ; and
• constructing the interface .
While these are conducted somewhat sequentially , there may be iteration as modification and refinement in one step impacts on the other two .
The model structure is the set of equations or behavior rules that fully specify how a change in one element of the model is related to the various factors that contribute to that change . That is , the model structure operationalizes the broad statements of the model design . It is determined through a combination of scientific knowledge , expert judgment , and data analysis .
For example , the model design may include a statement that ‘ If Z is large , then an increase in X will cause an increase in Y ’. Issues to be operationalized include : How are
67