Reconsidering the use of t-statistics in dendroprovenancing

Abstract

The t-statistic is a well-established parameter widely used in dendrochronology for cross-dating and, at least in preliminary stages, for dendroprovenancing. However, its application has increasingly been questioned due to its volatile validity and lack of clear statistical validation. This has prompted the adoption of alternative methods, though their implementation in dendroarchaeological contexts often presents challenges. This note aims to clarify key aspects of the t-statistic and its associated Pearson’s correlation coefficient in the context of dendroprovenance. In previous work, the authors demonstrated that, within the same species, a strong relationship exists between series correlation and geographic distance: higher correlations typically correspond to shorter distances between tree growth sites. Here, the focus shifts from the correlation coefficient to the t-statistic, highlighting both its potential and its limitations. Unlike the linear relationship observed between distance and Pearson’s correlation coefficient, the association between distance and the t-statistic is more pronounced at shorter distances, where correlations are higher. Additionally, the overlap between the tree-ring series has important implications in dendroprovenance analyses as a greater overlap facilitates high t_BP values by making these values less informative.