Statistical analysis of sulphide inclusions in diamonds: A contribution tothe multivariate analysis of compositional data

The geochemistry of sulphide inclusions in southern African diamonds (kimbe rlite occurrences designated as Roberts Victor, Koffiefontein, Jagersfontei n and Orapa) is analysed by multivariate statistical methods that take into account the compositional nature of the data. The data also encompass delt a(13)C determinations on the host diamond. Fe and S are strongly associated in all inclusions, as to be expected from the dominance of those two eleme nts in the material. The rank correlation coefficient suggests the carbon i sotope of the host and Ni of the inclusions to be significantly associated in two samples, notably Koffiefontein (negative relationship) and Jagersfon tein (positive relationship). The ordination yielded by constrained canonic al variate analysis indicates that the chemical values serve to give each l ocality a separate identity. However, the Botswanan Orapa data is bipartite . The differentiation of localities reflects the mineralogical status of th e inclusion-sulphides. Ordinations based on mss (= monosulphide solid solut ion) samples alone indicate that the evidence for heterogeneity manifested in the full set of data also appears in the subsample. The cloven nature of the Orapa data emerges also for the mss data-set. A brief orientation in t he covariance structure of compositional data is provided in an appendix.