��������������������������
arises from elements of the modeling process , as well as characteristics of the particular model . Thus , policy analysts should understand the process of modeling so as to be most effectively involved and maximize the usefulness of the final model .
Feasibility , the final consideration of model quality , considers practical constraints such as time , cost , data needs , specialist modeling skills , and other resources . In practice , all modeling techniques require relevant modeling expertise and similar types of resources . Additional time and funding allows better models to be developed or further detail to be included , but do not substantially impact on the modeling techniques that can be used . It is therefore difficult to specify general rules about resource needs for different modeling techniques ( though guides such as ( RIGHT , 2009 ) provide some insight ).
The key aspect of feasibility for which the needs of different modeling techniques vary widely is data . Diagrams that capture people ’ s perspectives on an issue do not need data other than those perspectives . In contrast , policy relevant mathematical or computer simulations typically require extensive quantitative data to parameterize the modeled relationships , though subject matter expertise may be substituted in some situations .
Feasibility is not explored in further detail because there are few implications for the choice of modeling technique . Relevant issues ( such as data requirements ) are revisited in the discussions of functionality and accuracy .
III - Functionality
Functionality concerns the reasons for
choosing modeling to examine a specific issue rather than , or in addition to , other methodologies such as literature analysis , interviews or statistical analysis . Many reasons for modeling have been identified ( Banks , 1999 ; Epstein , 2008 ; Greenberger et al ., 1976 ; Hodges & Dewer , 1992 ; Kelly ( Letcher ) et al ., 2013 ; Sternman , 1991 ). For dealing with policy issues , there are three broad domains of functions , which are particularly important : knowledge synthesis , management of unknowns , and policy support ( Bammer , 2013 ). Models can assist with specific tasks within each of these domains .
In practice , much of the functionality arises from the modeling process rather than simply the model . Therefore the process is described first , followed by the discussion of the functions supported by models and how they are delivered .
A . Th������������������
There are many different modeling
techniques . Some require mathematics and computers , while others need only pen and paper . Regardless , they all involve a development process , and a good model depends on undertaking that process effectively .
The process comprises four broad phases : Design , Build , Test , and Use . While the general flow is from Design to Use , there is substantial iteration between the Build and Test phases to ensure the model is constructed appropriately . For example , the building process can reveal gaps in the design , and testing a prototype can expose flaws . Even if there are no design flaws , the design could be changed because initial end user exposure to the model stimulates ideas and leads them to suggest refinements . Thus , the four phases should be interpreted as a guide rather than a prescriptive and ordered process .
Within these four phases there are several steps ( Banks , 1999 ; Jakeman et al .,
65