Univariate and multivariate outlier detection in exploration geochemistry: A case study of Sari Gunay epithermal gold deposit

Outliers are very important in exploration geochemistry and they can greatly impact the results of the statistical methods especially separation of anomaly from background. Thus, recognition and decision about removal or correction of them is one of the first steps in the analysis of geochemical data processing. Outliers can be identified from a univariate, bivariate, or multivariate perspective based on the number of variables considered, however the bivariate detection is not the aim of this study. Since geochemical data are compositional data, meaning that they represent a closed number system, an appropriate transformation should be used prior to any analysis. The most widely used remedy for this issue is the family of logratio transformation among which the isometric was selected for this study. The data were transformed using the isometric logratio transformation and then boxplot and robust Mahalanobis distance were applied for univariate and multivariate outlier identification respectively. The threshold value for univariate method was the upper inner fence (third quartile plus 1.5IQR), while for Mahalanobis distance an adjusted quantile, deviation of the empirical distribution function of the robust Mahalanobis distance from the theoretical distribution function, was used as the cut off value for identifying outliers. It was demonstrated that regarding the multivariate nature of exploration geochemical data, using multivariate methods for outlier detection is more accurate.