Finally, a three-dimensional plot of PC1 versus PC2 versus PC3
has been constructed, which is shown in Fig. 8, to give a better view
of the groups' distribution, showing that seven groups are formed.
The information obtained showed that two regions (CCS and SCa)
have a very similar mineral profile. The other regions (VVi, VSF, OPR,
CGa, SGa and PBa) can be successfully separated, by evaluating their
mineral composition and applying a PCA. To confirm the information
obtained by the PCA a HCA dendrogram has been constructed using
the Ward's method and the auto-scale data, which is shown in Fig. 9.
The dendrogram shows that one sample from SGa is similar to the
PBa samples, which can also be confirmed in the 3D PCA of Fig. 8,
but the other SGa samples do not show the same similarity, so that
two different groups have been formed. This can be explained by the
proximity of the regions PBa and SGa and the similarity in climate
and soil composition. With that it is possible to observe the
formation of seven different groups with high similarity index,
confirming the information obtained by the PCA. So, according to
the mineral composition it is possible to differentiate Brazilian wines
by their origin using only the information about the content of four
elements. The software used for all the statistical evaluation was
Matlab 7.11.0/PLS_Toolbox 6.2.1.