3.2. Chemometric analysis of NMR dataset
Fig. 2 exhibits some results obtained from the exploratory
principal components analysis. Regarding the variations covered by
the first two principal components (PC1 and PC2), one can see from
Fig. 2(a) that PC1 explains most of the variation in the data (96.8%);
it indicates the high level of dependence between the bins as well
as the goodness-of-fit of PC1 to capture such dependence. In turn,
two samples from the non-irradiated group were considered
Fig. 1. Typical 1H NMR spectrum (300 MHz) of chloroform-d extract of soybeans.
268 A.S. Ribeiro et al. / Food Control 36 (2014) 266e272
outliers and had to be removed from the analysis. Further, from the
PCA scores scatter plot involving PC1 and PC2, it can be seen that
though one cannot discriminate between different levels of gamma
irradiation, but it is clearly possible to distinguish between irradiated and non-irradiated soybeans. The PCA scores scatter plot using
the three fi rst principal components also indicates be possible to
separate the irradiated and non-irradiated samples only (data not
shown). Thus, considering that the irradiated samples form a
unique cluster, we grouped them into the namely irradiated samples, for the PL S-DA analysis