Year: 2006
Type: dissertation
Abstract: Statistical models for the calibration of both independent and related groups of \nradiocarbon determinations are now well established and there exists a number of \nsoftware packages such as BCal, OxCal and CALIB that can perform the necessary \ncalculations to implement them. When devising new statistical models it is important \nto understand the motivations and needs of the archaeologists. When researchers select \nsamples for radiocarbon dating, they are often not interested in when a specific plant or \nanimal died. Instead, they want to use the radiocarbon evidence to help them to learn \nabout the dates of other events, which cannot be dated directly but which are of greater \nhistorical or archaeological significance (e.g. the founding of a site). \nOur initial research focuses on formulating prior distributions that reliably represent \na priori information relating to the rate of deposition of dateable material within an \narchaeological time period or phase. In archaeology, a phase is defined to be a collection \nof excavated material (context or layers) bounded early and late by events that are of \narchaeological importance. Current software for estimating boundary dates only allows \nfor one possible type of a priori distribution, which assumes that material suitable for \ndating was deposited at a uniform rate between the start and end points of the phase. \nAlthough this model has been useful for many real problems, researchers have become \nincreasingly aware of its limitations. We therefore propose a family of alternative prior \nmodels (with properties tailored to particular problems within archaeological research) \nwhich includes the uniform as a special case and allows for more realistic and robust \nmodelling of the deposition process. We illustrate, via two case studies, the difference in \narchaeological conclusions drawn from the data when implementing both uniform and \nnon-uniform prior deposition models. \nThe second area of research, that we take the first steps towards tackling, is spatiotemporal \nmodelling of archaeological calibration problems. This area of research is \nof particular interest to those studying the response of plants and animals, including \nhumans, to climate change. In archaeological problems our temporal information \ntypically arises from radiocarbon dating, which leads to estimated rather than exactly \nknown calendar dates. Many of these problems have some form of spatial structure yet \nit is very rare that the spatial structure is formally accounted for. The combination of \ntemporal uncertainty and spatial structure means that we cannot use standard models \nto tackle archaeological problems of this kind. Alongside this, our knowledge of past \nlandscapes is generally very poor as they were often very different from modern ones; \nthis limits the amount of spatial detail that can be included in the modelling. \nIn this thesis we aim to make reliable inferences in spatio-temporal problems by carefully \ndevising a model that takes account of the temporal uncertainty as well as incorporating \nspatial structure, to provide probabilistic solutions to the questions posed. We illustrate \nthe properties of both the conventional models and the spatio-temporal models using a \ncase study relating to the radiocarbon evidence for the Late glacial reoccupation of NW \nEurope. \n
Author: Angela Jane Karlsberg
Language: English
Subfield: Archeology
Field: Social Sciences
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W2420424078