IBMs contain much more information about the state of all the individuals and their environment than we can process. We thus need to aggregate this information into one or more aggregated system-level state variables. We need to define currencies (often also referred to as 'indices') that allow us to characterize a single run of the model over a certain time horizon by, ideally, one single number. This number is then used as a currency to compare different parametrizations or formulations of the model. For example, in models addressing extinction risk of small populations, the risk of extinction over a certain time horizon is such a currency; or, if we want to compare model output to an observed time series, we can use root mean square deviation as a currency. In general, any quantitative measure of how good a certain observed pattern is reproduced is a good currency. Often, it is not clear from the outset which currency allows for deepest insights, so currencies are experimental and have to be tested and usually we will need to consider a set of currencies simultaneously.
Was this article helpful?