End-to-End Bayesian Inference for Summarizing Sets of Radiocarbon Dates

Summary

This is an abstract from the "Constructing Chronologies II: The Big Picture with Bayes and Beyond" session, at the 86th annual meeting of the Society for American Archaeology.

Aggregations of radiocarbon 14C dates are seeing increasing use as proxies for the relative population size through time of past societies and regions. Two major problems complicate the use of sets of radiocarbon dates as demographic proxies: the bias problem and the summary problem. The bias problem exists because the radiocarbon dates available for study are not necessarily representative of past population sizes (for example, more research effort has been allocated to one era over another). The summary problem exists because of sample size limitations, uncertainty in radiocarbon determinations, and ambiguity due to the radiocarbon calibration curve. The focus of this presentation is a novel statistical method for solving the summary problem, end-to-end Bayesian inference. I demonstrate the superior statistical and empirical properties of this approach compared to the current, dominant approach for summarizing sets of radiocarbon dates, summed probability densities (SPDs). I conclude by discussing approaches for addressing the bias problem, notably the need to fuse multiple types of data to improve demographic inference. Pertinent data include skeletal data on age-at-death, health, and stable isotopes (informative of migration); ancient and modern genetic data; radiocarbon dates; house and pottery counts, etc.

This Resource is Part of the Following Collections

• Constructing Chronologies II: The Big Picture with Bayes and Beyond   •
• Society for American Archaeology 86th Annual Meeting, Online (2021)

Keywords

Record Identifiers

Abstract Id(s): 32599