In order to determine which variables are the most involved in
the clustering of the samples on the score plot, we can represent
the variables projection (Fig. 5) on the PC1 and PC2 space.
A variable is considered as influent if its norm projection on the
PC axis is high. For example, the fitting parameters of bands 8 and
10 have an important negative contribution to PC1, and the fitting
parameters of band 7 have an important positive contribution to
PC1. This can explain for example the distance between the oils
and the gums on one side and the African copals and pine resins
on the other side. Indeed, bands 8 and 10 are absent from oils and
gums spectra and band 7 is present (Table1), whereas it is the
opposite for pine resins and African copals. A deeper look at this