New approaches to the characterization and classification of ceramics on the basis of their elemental composition

This study is a part of a larger investigation related to questions of the systemization of neutron activation analysis results. It considers the analysis of experimental data and focuses on the application of a multivariate statistical technique to the characterization and classification of ceramic artifacts in order to establish and assess their provenance. Central to the statistical methodology reported here is the determination of a specific characterization-classification function. This function provides a method of defining a group of archaeological ceramics on the basis of their elemental composition patterns. Conditional and expected probabilities calculated from the function are then used as a means of classifying unknown observations, indicating outliers, and detecting the existence of another group as yet unidentified. The utility of the function is demonstrated in a specific test case. The possibility of using the characterization-classification function to increase the portability of information and to reduce the need for centralized computer data banks is pointed out.