Abstract

Co-author: Robert J Clegg (University of Birmingham)

Using some examples I will show that simpler models can be less general and that complex models can be less realistic. It is therefore important to vary the complexity of a model to test the structural robustness of models, not just checking off the parameter sensitivity box. For example, one ought to test which processes (e.g. growth, diffusion, migration, predation, ) need to be included in a model, but having to spend a lot of time implementing further processes that may turn out not to matter means that this is often not done, especially towards the end of a project. Open-source, individual-based modelling platforms can help here if many groups contribute by implementing further processes enabling the user to quickly try out a bunch of processes. In the end, models are more useful if they are less wrong.