Exploratory compositional data analysis

Peer reviewed paper. This paper presents the first steps that should be performed whenever the study of a compositional dataset is initiated. Centre, variation matrix and total variance of a compositional dataset are introduced. In addition the biplots are also introduced as a powerful tool to analyse and discover special features related to subcompositions. The exploratory methodology is applied to a dataset consisting of the major, minor and trace elements composition of soil samples from several places in Tuscany (Italy). The structure of the data, collected from three different known country rocks of ophiolitic nature (basic and ultrabasic rocks), represent an interesting case study for experimenting on new methodologies of statistical investigation and for pointing out differences related to parental chemistry and mineralogy as well as the nature of processes to be related to the subsequent evolution.