Cluster analysis is a multivariate statistical technique that may be used to elaborate on sample and element associations when a wide range of elements or characteristics are determined from geochemical samples. It relies on an object or sample being defined by a number of attributes or variables which may be quantified- or numberically coded. A data matrix is therefore obtained consisting of a number of samples with their corresponding set of suitable coded or quantified characters. Similarity coefficients are calculated between each pair of samples or variables depending on whether samples (Q—mode) or variables (R-mode) are being clustered. Most clustering methods consist of grouping the samples or variables on the basis of the computed similarity coefficient, the nucleus of clusters being formed by joining the samples with highest similarity coefficient and gradually admitting more as the similarity coefficient is lowered. Other clusters are eventually initiated until finally all the samples are linked. Various options of clustering methods exist, depending on the criterion of entry of a sample or element into a cluster, e.g. weighted and unweighted pair-group average linkage method. The end product of the clustering procedure can be summarized as two-dimensional hierarchical diagrams called dendrograms.