Comparison of log-ratio and log10 chemical elemental data analysis of Central Amazonian pottery and archaeological implications

Abstract

The additive log-ratio (alr) transformation is recommended as one of the most robust data transformations for multivariate analysis of archaeometric compositional data. However, alr and other transformations are not mutually exclusive and can be combined to assess different aspects of an archaeometric data set, such as the addition of temper, post-depositional effects in pottery and associated archaeological implications. This study presents a comparative analysis of a multi-element data set of pottery from Lago Grande and Osvaldo archaeological sites in the Central Amazon, which are considered a microcosm of the region. The concentrations of nine chemical elements (La, Lu, Yb, Ce, Cr, Eu, Fe, Sc, and Th) measured by instrumental neutron activation analysis (INAA) were subjected to alr transformation, prior to chemical fingerprinting by cluster analysis (CA) and principal component analysis (PCA). The results were compared to a previous work using the log10 transformation. Multivariate analysis of variance (MANOVA) was employed to test for statistical differences between the chemical groups, and self-organizing maps (SOMs), a type of artificial neural network, were used for comparison due to their advantage of not depending on any specific data distribution assumption. In general, the results suggest the existence of socio-cultural interactions between Lago Grande and Osvaldo, which could have occurred through trade, exogamic marriage and territory sharing. In a broader perspective, the exchange networks corroborated by the results favor theories that minimize the role of ecological constraints in the emergence of social complexity and sedentary occupations in the Amazon region.