FAQ: How do you know when your model is sufficiently complex?
Submitted by: Steve Waydo
Submitted on: October 8, 2003
Knowing when your model is sufficiently complex is often a somewhat subjective judgement call and one that requires some experience to be comfortable with. Some choices are clear, for example in the model of Bob's interaction with the terrain from class we certainly don't have to worry about the motion of a person riding within the vehicle. Others are less so, for example whether a one-wheel model of Bob is sufficient or if it would miss important features captured by the two-wheel model.
When we're really not sure what we end up doing is building models at multiple levels of complexity and comparing the results - if we weren't convinced that a single-wheel model of Bob was sufficient for the problem we discussed in class, the fact that the results were almost identical to those obtained from the more complex two-wheel model should convince us. Often a simple model will be used for most of the work and occasionally verified against a more complex model that would be more difficult to work with most of the time.
As was implied by the feedback loop in the design path put on the board in class, sometimes we discover our model was not sufficiently complex only when we build and test a system and it does not behave as expected. At that point we try to identify the neglected factors that were important and modify our model to include them.