M. Járó - L. Költő szerk.: Archaeometrical research in Hungary (Budapest, 1988)
Analysis - BALLA Márta, BÉRCZI János, KEÖMLEY Gábor, ROSNER Gyula, GABLER Dénes: Provenance studies of ceramics by neutron actiwtion analysis
11/2 xijc denotes the standardized concentration of the kth element measured on sample i xj,k denotes the standardized concentration of the kth element measured on sample j The correlation coefficient (rj j) shows whether the tendency of elementary concentration changes between two samples is of similar type or not: n 2 -2 fc 2 x^-nxi 1/2 n 2 -2 k=l J * J 1/2 The greater value of i[ j marks the greater similarity between two samples. Samples with the highest mutual similarity are selected on the basis of the computed values and calculations are repeated in order to construct groups (clusters) diverging from the initial pair or pairs step by step. Eventually a system could be developed — structurally like a branch of a tree - characterizing and illustrating the similarity relations between the samples. Initially the "Cluster" 1 program was run on the Technical University's R32 type computer and later on a Commodore 64 microcomputer. The other approach - named similarity analysis - was applied when sherds of unknown origin were to be placed into one of the archaeologically well defined workshops. In this case the first step is to calculate die average concentrations of the elements determined and the standard deviation of the measured data concerning each workshop. Then elementary concentratior values of the sample of unknown origin are compared with the characteristic values of the different groups one by one and the probability of belonging to a group is given by each element. Knowing these probabilities one could give the chance — based on a given element — whether the sample belongs to a workshop or not. Having multiplied the values characterizing these "chances",, a number is obtained which is already significant for the probability of belonging to a given group conœrning every investigated element. This calculation is carried out for all the probable workshops, the ratio of the values obtained definitely places the sample into one of the groups is most of the cases. In its archaeological meaning, any arbitrary number of workshops and samples to be identified can be classified by the developed computer program. Error of measured data, accuracy, reproducibility If the errors made during measurements are known it is possible to calculate the error of measuring the concentration. We give here — as the ntinimum error of the measuring process the measurable error obtained during the peak area calculation as a result of the statistical behaviour of radioactive decay (per cent, relative (Table 1)).
