Vörös A. szerk.: Fragmenta Mineralogica Et Palaentologica 14. 1989. (Budapest, 1989)

concerning major element composition data. As the eigenvalues show in Table 2

f the first two eigenvectors extracted from variance-covariance matrix have higher information content (88. 7%) than the first two eigenvectors extracted from the correlation matrix. The principal component variation diagrams can reveal the characteristic differentiation trend(s) within a volcanic sequence. As LE MAITRE (1968") pointed out, the curvature of trends as well as the scatter of the points about the main trend is most marked in the alkali provinces, while an almost linear trend indicates sub-alkali sequence. The F1-F2 plot of Mecsek volcanics (Fig. 5) indicate two different chemical evolution processes, ending with phonolites and trachytes respectively. This is corresponding with the result of NLM. Note that the F1-F2 plot extracted from the variance-covariance matrix is more expressive than the plot using correlation matrix (Fig. 5). Performing PCA on the subsets of the volcanics (step-wise PCA) it can reveal the components responsible for the changes in composition of the clusters of the rock types. A detailed example for this technique is included in UPADHYAYA et al. (1988). Table 3 Eigenvectors and eigenvalues of Mecsek volcanics in = 349 samples) Eigenvectors : Eigenvalues : 1 2 3 4 Si0

2 -0 595 0 457 0 447 -0.018 Ti0

2 0 721 -0 091 -0 120 0.041 A1

2

0

3 -0 311 0 070 0 386 -0.648 Fe

2

0

3 0 012 -0 780 -0 078 0. 185 FeO 0 755 0 491 '1 158 0. 128 MnO 0 018 0 026 -0 064 0. 070 MgO 0 827 -0 267 -0 001 0. 160 CaO 0 369 0 084 -0 819 0. 224 Na

2

0 -0 404 0 570 0 2 51 -0.463 K

2

0 -0 688 0 203 0 315 0. 161 -0 092 -0 018 0 008 0.816 co

2 -0 045 0 079 -0 904 0. 044 H

2

0+ 0 066 -0 741 0 215 -0.028 H

2

00 175 -0 711 0 037 -0.0 56 3 042 :? 577 2 098 1. 465 Applying PCA for the whole data containing both fresh and altered samples, all the processes affected on the composition can be revealed. PCA for 349 samples of Mecsek volcanics were carried out. The loadings of the first four eigenvectors are shown in Table 3. Three essential processes affected the chemistry of the volcanics. The most significant process was the fractional crystallization, indicated by the high absolute values of MgO, FeO, Ti0

2 and Si0

2

, K

2

0. The second one may have been the oxidation, the submarine alteration and weathering (Fe

2

Og, +H

2

0, -H

2

0), while the third one was the CaCOß-enrichment (CaO, C0

2 with high negative values). CONCLUSION The application of multivariate mathematical methods proved to be effective on the petrochemical study of a volcanic suite. Classification of the main rock types can be achieved by the use of cluster analysis. Standardization of the data are not recommended. If it is possible, both the Euclidean distance and cosine theta (or theta) coefficient can be used as a similarity measure, while among the linkage methods the weighted-pair group average is the