The statistical analysis of geochemical data employs the main statistical techniques of averaging, probability distributions, correlation, regression, multivariate analysis and discriminant analysis. A particular problem with major element geochemical data is that it is constrained; that is, the compositions sum to 100% and the data are ‘closed’. A related problem arises when ternary plots are used to display geochemical data. Techniques are described to accommodate the problems associated with compositional data which include log-ratio conversions and the biplot diagram. Further statistical problems arise in the area of ratio correlation as advocated in Pearce element ratio diagrams, which is not recommended. Applications to trace elements and radiogenic isotope correlations are discussed. The details of discriminant analysis are outlined as a prelude to a more detailed discussion of tectonic discrimination diagrams considered in Chapter 5.