On the use of discriminant analysis techniques for classifying chemical data from panned heavy—mineral concentrates—Central East Greenland

Multi-element analyses of more than 600 panned heavy-mineral concentrate samples from the Jameson Land area of central East Greenland were investigated by discriminant analysis which, combined with an a priori knowledge of the geology, was used to assist interpretation and classification the data. The importance of sample density, and the method and degree of sample grouping were investigated. Also the effect of using either the within-covariance or the pooled covariance matrices in the discriminant functions was studied.Discriminant analysis assumes the element concentrations to be normally distributed and that equality of within-covariances exists. These assumptions were not always met, and this was probably the main reason why one-third of the samples could not be classified. However, the study shows that interpretation of the data, although it is historical and forms only part of a regional reconnaissance survey, benefits from the application of discriminant analysis.