The obtained model not only represented excellent prediction
ability for training samples, but also it could predict correctly the
class membership of all samples in the test set. The non-error rate
(NER,% of correctly assigned samples) for calibration and prediction
samples are shown in Table 1, while in Table 2 specificity
and sensitivity achieved on all samples are shown, respectively.
As we can see from these Tables, iPLS-DA with extended multiplicative
scatter signal correction preprocessing and using the 10 factors
gave NER equal to 42 and 46% for the calibration and the test
samples, respectively, whereas iECVA showed the best classification
results and performed better than PLS-DA and iPLS-DA. The
same conclusions can be reached by looking at Table 2 since iECVA
consistently yields perfect values of both specificity and sensitivity
on the calibration and prediction samples with respect to the other
employed classification methods.