Data selected from an extensive major element database of Cenozoic volcanic rocks (including calc-alkaline andesites, dacites, rhyolites, and alkali basalts) of Hungary are used to illustrate the detection and modeling of subcompositional patterns using a statistical analysis based on the assumption that relative differences between the observed values are more meaningful than absolute ones. In particular, two roughly linear compositional patterns (associated one to the alkaline basalts, the other to the calc-alkaline series) are revealed and evaluated, and it is shown how principal component analysis can be used to obtain the estimated subcomposition of their incidental intersection point.