Hierarchical Factor Classification of Dendrochronological Time-Series

Abstract

In this paper, Hierarchical Factor Classification (HFC), an exploratory method of classification of characters is introduced, in comparison with Principal Component Analysis (PCA) in order to show its advantages, in particular when dealing with time series. Exploratory data analysis may play a very relevant role in the understanding of the structure of a data set prior the use of statistical methods – as hypothesis testing and inference, and models. The study of tree-rings time series through exploratory methods may also take advantages, by allowing some interpretation to be further checked via a small number of statistical tests. In particular, while providing overall results close to those of PCA, HFC complements it, by providing a classification of the time-series and estimating a representative chronology for each group, common to the clustered ones. As case study, a data set is taken from literature, composed by five synchronous 79 years-long chronologies of Pinus pinea L., from five different populations scattered along the Tyrrhenian coast in peninsular Italy. HFC suggests how conveniently aggregate the chronologies, by showing similarities and differences between them, otherwise unnoticed, suggesting to limit the aggregation to three chronologies only.