Models are connected to our theoretical understanding of system operation via a set of morphisms or relationships that link the model and represented system (Zeigler et al. 2000). The aim of model assessment is to establish the strength of these morphisms via a process of model verification, validation and a more general critique of the entire modeling process. This section highlights a portion of the assessment carried out on this model.
The first step in the model assessment process we refer to as model verification and is concerned with ensuring that the model has been built correctly relative to the model design document (Zeigler et al. 2000; Balci 1994). The primary aim of this step is to eliminate implementation errors. This involves checking the actual implementation against the design specification, a summary of which is given in Appendix A. In addition, the model was constructed to facilitate highly detailed tracking of individuals. Each time the crab is updated (on average once per hour) all of a crab's state variables and its local environmental variables can be exported. This information enabled us to analyze whether or not the rules specified in the model design were actually occurring in the situations they were supposed to. Both these steps formed the basis of the initial model assessment, but we have included little of this low level assessment here because it was solely concerned with ensuring that the model accurately represents the model design document (Appendix A). Thus, each low-level component of the model has been assessed and is functioning as described.
The next step in assessment is model validation and is concerned with building the right model. This involves substantiating that the behavior of the model is ``close enough'' to the behavior of the source system so that it is impossible to distinguish the behavior of the systems in the applicable experimental frames (Zeigler et al. 2000; Balci 1994). An experimental frame is a specification of the conditions under which both the source system and model are to be observed or experimented with. Specifying the experimental frames ensures a fair comparison between the output of the two systems. Examples of experimental frames are given below. The concept of an experimental frame extends to situations where little empirical information is available. The literature can be consulted for further guidance (e.g., Kleijnen 1995; Balci 1997; Zeigler et al. 2000; Balci 1994; Rykiel Jr. 1996; Yilmaz and Balci 1997). At this stage, assessment of emergent properties such as the time it takes a crab to reach sexual maturity are of particular importance since such times are not hard-wired into the model, but emerge based on the interaction of a crab's development and energy balance rules in conjunction with the environment it encounters.
The larger aim in documenting the model implementation and assessment is to facilitate the larger process of model critique. This allows others (especially those not involved in developing the model) to understand the modeling decisions made and their implications.