3.2. Principal component analysis
Principal component analysis is utilized to provide an overview
of the capacity to characterize vegetable oils based on gas chromatography
data. By choosing the first three principal components
(PC), 98.47% of the total variability is explained (PC1 = 54.76%,
PC2 = 30.55%, PC3 = 13.16%) indicating that we do not lose a considerable
amount of information by keeping only the first three
PCs. The scores for the first three PC are plotted as a scatter diagram
in Fig. 2. Generally, the stereochemical distribution of the
mixtures depends on the nature of the oils. It is clear that several
clusters are formed, corresponding to the five different kinds of
mixture. In each cluster, the samples are dispersed from the center
(six pure sesame oils) to the rim (the five kinds of pure non-sesameoil). This may be explained by the fact that similarity of samples is
determined on the basis of the measure of correlation coefficient in
PCA analysis, and the mixtures of specific two oils is of course
highly correlated. At the same time, the dispersion direction of
each cluster along PC1, PC2 and PC3 is caused by the different
kinds of oils mixed with the pure sesame oil. Moreover, it seems
that all clusters have the common center, and the higher level of
a non-sesame oil contained in a mixed sample, the further distance
of this sample from the pure sesame oil.