Left: bi-plot after Principal components and classification analysis based on correlation of the content of the 54 chemical elements (empty marker, not all labelled) in the 29 soil samples (full marker).

<p>The variance explained by the each principal component is shown in parentheses. Right: plot of canonical scores after Discriminant analysis with canonical analysis performed with the four chemical elements most, and significantly, correlated (absolute value) with factor 1 (Cs, Ga, Ca, S) and factor 2 (Cd, Zn, Hg, Sr). Classification of soil samples has an high probability to be correct (93%) as the predicted classification deviates only for sample “Le”, that is not distinct from “X” (see <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0020222#pone-0020222-g001" target="_blank">fig. 1d</a>). Since these two samples are in the same group (the green one), group classification is 100% correct. For technical details of multivariate techniques see Statistica 7.0 Electronic Manual (Statsoft Inc., 2005).