Identification of volatile compounds was achieved by comparing
the extensive information obtained. All spectra were aligned
and normalized to the total peak intensity. To quantify volatile
compounds, the samples were tested in triplicate, and integration
areas based on the total ion chromatogram were normalized to the
areas of the internal standard and averaged. To determine the
reproducibility of triplicate injections and identify peaks carried
across samples, two-way analysis of variance (ANOVA) (sample,
injection) was performed for each volatile compound peak. All
peaks varied significantly across samples, with only three peaks
varying significantly across replications. Principal component
analysis (PCA), a multivariate pattern recognition procedure, and
the unsupervised clustering method, which reduces the
dimensionality of multivariate data while preserving most of the
variance therein, were utilized on the volatile fraction fingerprints
of the investigated samples to reveal the influence of Se-rich operation,
storage temperature, and time on the discrimination of the
first two principal components.