| Popis: |
Recent years have seen an increasing use of quantitative approaches in Roman studies. The growing amount of openly available archaeological datasets provides a wealth of information that could be used to inform current debates. One of the most interesting aspects of this Open Data approach to archaeology is the capability of integrating evidence collected across thousands of different sites. This trend can be very useful to understand large-scale patterns such as trade and market structure, but are more data enough? The newly available datasets will be useful if we can use them to assess the plausibility of our ideas about the economic dynamics of the Roman Empire. This comparison between working hypotheses and data is particularly challenging in the Humanities due to the fact that ideas are traditionally expressed as natural language instead of mathematics. This difference between the language used to define evidence and ideas does not allow us to quantify the error between collected data and what our explanations would predict, thus inhibiting our ability to directly assess and compare competing hypotheses. In other words, if we do not translate our explanations into formal models then it will be very difficult to know if we are wrong, and no amount of data will help us decide. We discuss here this challenge by exploring two different hypothesis-testing frameworks in the context of Roman studies: (a) null-hypothesis significance framework and (b) model selection. The benefits and limitations of both systems are discussed through a range of case studies linked to the production and distribution of olive oil. Results suggest that the combination of new data and hypothesis-testing methods can help us inform traditional debates of Roman studies as long as these quantitative approaches are able to tackle the complexities of archaeological research, including multiscalar dynamics, data fragmentation or high levels of uncertainty. |
