Graphical approaches to multivariate data analysis using archaeological data

Bibby, K.J., 1997. Graphical approaches to multivariate data analysis using archaeological data. MPhil, Nottingham Trent University.

10183549.pdf - Published version

Download (17MB) | Preview


This thesis investigates the application of some recently developed statistical methodology to problems arising in the analysis of multivariate archaeological data. Specifically, data sets on the chemical composition of glass fragments found in archaeological contexts are used.

The statistical methods often used to investigate such data (for the existence of groups, etc.) are sensitive to the presence of outliers. One focus of the thesis is a comparison of the performance of different methods of outlier detection with such data, including recently developed methodology.

Removal of outliers makes it easier to detect patterns in the data. Standard methods such as principal components analysis and cluster analysis are used for this purpose. Results are displayed using kernel density estimates (KDE's) in a variety of ways. Although KDE's are now an established statistical technique, their application to archaeological problems is comparatively novel.

For the data sets used here the newer outlier detection methods usually differed little from the methodologies they are supposed to improve on, in terms of outliers detected. They are also not ideally suited to data sets having the kind of structure exhibited by those used. The ECDE's proved valuable in displaying structure in the data, and it was often possible to provide a substantive explanation for the structure in terms of glass chemistry and colour. It was also observed that the statistical outliers detected could, in retrospect, be recognised to be archaeologically or physically unusual with respect to colour.

The substantive analyses raise a number of interesting questions. For example, observed structure may be related to quite subtle colour difference (reflected in the chemistry), but colour is often not recorded. Even where it is, sample sizes smaller than those used here may not allow the detection of structure. Finally, one analysis revealed structure that will require further archaeological investigation.

Item Type: Thesis
Creators: Bibby, K.J.
Date: 1997
ISBN: 9781369317138
Divisions: Schools > School of Science and Technology
Record created by: Jeremy Silvester
Date Added: 01 Oct 2020 13:38
Last Modified: 20 Sep 2023 10:18

Actions (login required)

Edit View Edit View


Views per month over past year


Downloads per month over past year