Statistical modelling in archaeology: some recent trends and future perspectives

This paper reviews the application of statistical models in archaeology in the last decade, focusing in particular on multilevel models, statistical treatment of missing data and measurement error, and simulation-based generative inference. These techniques are designed to 1) account for the nested and hierarchical nature of the archaeological record, 2) formally integrate different forms of data uncertainties, and 3) provide a more direct inferential link between formal theory and observational data. The extent to which archaeology has engaged with these methods is variable, but it can be argued that none are currently regarded as part of the standard analytical toolkit in quantitative archaeology. The objective of this paper is to promote awareness of the existence of these techniques and highlight the consequences of ignoring the underlying problems that these statistical methods can address. • Multilevel models are suited to handle sample imbalance and interdependence commonly found in synthetic datasets. • Ignoring missing data and measurement error can lead to biased inference. • Statistical approaches for handling missing data and measurement error exist and should be used more routinely. • Simulation-based generative inference can offer a bridge between theory and data.