Troubleshooting

Evaluating Model Performance

The principal products of a HexSim simulation are population trends, population distributions, and trait distributions for each population. These simulation products integrate all of the parameter values and interactions built into a scenario. If the model design captures the critical life history traits, stressors, and other features, and if it has been carefully calibrated using available data sets, then the simulation results should be meaningful. Still, mistakes can easily be made, and plausibility testing should be employed to help identify errors. This section briefly discusses plausibility testing in the context of the simulation products mentioned above.

Model Calibration

This section presents some approaches that may be useful for constructing well-calibrated HexSim scenarios that produce plausible results. This is not an exhaustive treatment, and this topic is not even an exact science. The most important thing users can do is to always construct the simplest plausible model. Adding complexity beyond that which is necessary to capture critical dynamics make models harder to understand and consequently less useful. However, it's often only through trial and error that we determine how much complexity is absolutely necessary.