An indicator of this phase is the increased use of quantitative analyses that focus on both a better mechanistic understanding of the model and on relating the model to real-world phenomena and mechanisms.
The price for this generality is that the models usually do not deliver testable predictions, and it remains unclear how well they really explain observed phenomena.ġ.3 The second phase in agent-based modelling appears to begin once a critical mass of models for certain classes of questions and systems has been developed, so that attention shifts from representation and demonstration to obtaining actual insights into how real systems are working. Most models developed in this phase are designed to demonstrate general mechanisms or provide generic insights. Typically, model evaluations are qualitative, and fitting to data is not a major issue.
The focus in this phase is usually more on how to build representations than on in-depth analyses of how the model systems actually work. First, most ABMs in a certain field of research are designed and analysed more or less ad hoc, reflecting the fact that experience using this tool must accumulate over time. 2011 Railsback & Grimm 2012).ġ.2 This establishment appears to have occurred in at least two phases. ABMs have therefore become an established tool in social, ecological and environmental sciences ( Gilbert 2007 Thiele et al. The use of ABMs is thus required for many, if not most, questions regarding social, ecological, or any other systems comprised of autonomous agents. ABMs are used when one or more of the following individual-level aspects are considered important for explaining system-level behaviour: heterogeneity among individuals, local interactions, and adaptive behaviour based on decision making ( Grimm 2008). Parameter Fitting, Sensitivity Analysis, Model Calibration, Agent-Based Model, Inverse Modeling, NetLogoġ.1 In agent-based models (ABMs), individual agents, which can be humans, institutions, or organisms, and their behaviours are represented explicitly. In this way, we hope to contribute to establishing an advanced culture of relating agent-based models to data and patterns observed in real systems and to foster rigorous and structured analyses of agent-based models. Our overall aim is to make agent-based modellers aware of existing methods and tools for parameter estimation and sensitivity analysis and to provide accessible tools for using these methods.
The Supplementary Material includes full, adaptable code samples for using the presented methods with R and NetLogo. We then list the packages in R that may be used for implementing the methods, provide short code examples demonstrating how the methods can be applied in R, and present and discuss the corresponding outputs.
We briefly introduce each method and provide references for further reading. Because NetLogo and R are widely used in agent-based modelling and for statistical analyses, we use a simple model implemented in NetLogo as an example, packages in R that implement the respective methods, and the RNetLogo package, which links R and NetLogo. To facilitate parameter estimation and sensitivity analysis for agent-based modellers, we show how to use a suite of important established methods. Both steps of the model development cycle require massive repetitions of simulation runs with varying parameter values. These insights foster the understanding of models and their use for theory development and applications. By exploring the sensitivity of model output to changes in parameters, we learn about the relative importance of the various mechanisms represented in the model and how robust the model output is to parameter uncertainty. Furthermore, sensitivity analysis is an important part of the development and analysis of any simulation model. Agent-based models are increasingly used to address questions regarding real-world phenomena and mechanisms therefore, the calibration of model parameters to certain data sets and patterns is often needed.