DATA ANALYSIS IN MATERIALS SCIENCE - THE PROVENANCE OF CERAMICS: A CASE STUDY

This thesis develops a basic methodology for dealing with the characterization and classification of complex materials systems on the basis of analytical results obtained from the experimental study of such materials. The situation addressed is one in which large sets of experimental data, expressed in terms of a number of interrelated variables, need to be evaluated on a level of complexity corresponding to the complexity of the problems being investigated. This task requires mathematical and statistical consideration not yet utilized in the field of materials science.;After reviewing the existing state of provenance studies, multivariate statistical analyses were employed first to define ceramic reference groups and then to classify the ceramic samples into previously defined and delineated groups. Functions were than arrived at for the characterization and classification of the samples.;Further development of these functions led to the formulation of: (a) a unique characterization-classification function which can serve as a "fingerprint" or a "signature" of a particular group of ceramics; (b) a procedure for the classification of samples into a group that has not been previously defined.;The theoretical approach developed for this thesis was tested on two sets of original archaeological data: ceramics from the Zagros region of the Near East; and ceramics from the Eastern Mediterranean (Israel and Cyprus). In both cases, the developed statistical approach yielded results that previous investigations could not obtain from the same experimental data.;A prototype problem that demands such data analysis is the assessment of the composition of archaeological ceramics in terms of their origin, i.e. provenance studies.;The methodology developed in this thesis could be applied to other materials science problems.