skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: The efficiencies and error-rates of Euclidean and Mahalanobis searches in hypergeometries of archaeological ceramic compositions

Technical Report ·
OSTI ID:6911035

The data presented permit a number of conclusions to be drawn: 1) In several well-characterized, tightly-knit groups, the use of 17 elements results, in Euclidean space, in essentially unique fingerprints. The error-rates are no more than one or two per ten thousand. Even when the elements are reduced to 10, there is still in these cases excellent characterization. When only 6 elements are employed, uniqueness is seriously compromised. Results based on less than elements are essentially meaningless. 2) In several other cases, there is a small error rate, typically 1-4%, in a Euclidean search utilizing even 17 elements. Short Euclidean distances cannot prove archaeological affiliation but, when combined with concordance of stylistic criteria, similarity in paste petrography, etc., may now give strong additional evidence as to long-distance contacts. It was suggested that the widespread occurrence of bivariate correlation in natural clays and archaeological ceramics could considerably sharpen the uniqueness of the ''fingerprint'' characterization of archaeological groups if one first transformed the space in such a way as to remove the correlations from the focal group. Those of us who have been involved in the characterization of archaeological ceramics through their analytical compositional signatures have had the intuitive feeling that the method worked fairly well; that one could distinguish sherds that lay in or out of particular clusters with some assurance, provided one employed a sufficient number of elements. The present paper demonstrates just how good that characterization can become, and roughly how many elements need be considered, to achieve a reliable result. 27 refs., 13 figs., 3 tabs.

Research Organization:
Brookhaven National Lab., Upton, NY (USA)
DOE Contract Number:
AC02-76CH00016
OSTI ID:
6911035
Report Number(s):
BNL-41458; ON: DE88013995
Resource Relation:
Other Information: Portions of this document are illegible in microfiche products
Country of Publication:
United States
Language:
English