This thesis concerns the importance of replication studies in agent-based modelling, specifically in the field of archaeology. As a case study, the MERCURY model by Brughmans and Poblome is...Show moreThis thesis concerns the importance of replication studies in agent-based modelling, specifically in the field of archaeology. As a case study, the MERCURY model by Brughmans and Poblome is replicated. In the first chapter, a background is given to ABM in general, as well as to replication and its importance and scarcity. Replication allows us to confirm the findings of existing ABM models, or reject them. The second chapter gives an abstract of Brughmans and Poblome’s research. It includes the archaeological background to their research, a precise description of the MERCURY model and a summary of Brughmans and Poblome’s conclusions. In chapter three, the process of replicating MERCURY is explained. Each version of the replication is described in great detail. This final version is statistically compared to the original model. The replication was found to be, for the most part, statistically equivalent to the original. The source of the despondencies between the various versions of the replication and the original model were due to inaccuracies in the description of the model as well as due to my own coding mistakes. Chapter four includes a brief discussion of existing criticism of the MERCURY model, as well as my own critiques. These critiques are mostly concern the details of the model and the way the authors interpreted their data. Additional experiments are performed to complement the experiments in the original study. I conclude that some of the issues I identify could weaken the original authors’ conclusions. The relation of my replication of MERCURY to other replication studies is discussed in chapter five. The final chapter my research questions are answered. I also shortly discuss how my experiences with replication could help future researches who want to publish agent-based models or replication studies.Show less
