BUCHI NIRCal - Manual (Page 205)

Chemometrics
NIRCal 5.5 Manual, Version A 205
3.18.91 V-Set Regression Intercept
Description Validation Set Regression Intercept of Original Property and Predicted
Property
Use To compare with the C-Set Regression Intercept
Method PCR / PLS
Matrices ID 28
Tip A value around zero is expected for a good calibration.
Details A better description of the V-Set regression coefficient.
Related Topic V-Set Regression Coefficient , V-Set Regression Slope
3.18.92 V-Set Regression Slope
Description Slope of the regression line equation of the validation set.
Use To compare with the C-Set Regression Slope
Method PCR / PLS
Matrices ID 29
Tip A value around one is expected for a good calibration.
Details A better description of the V-Set regression coefficient.
Related Topic V-Set Regression Coefficient , V-Set Regression Intercept
3.18.93 V-Set SEE (SEP)
Description Standard Error of Prediction: Standard Deviation of the Property Residuum of
the V-Set Spectra.
Use To compare with the C-Set SEE (SEC)
Method PCR / PLS (only with VS mode)
Matrices ID 25
Tip It should be as small as possible, but comparable with the standard deviation
of the lab method.
Details Precision. Available in the calibration protocol.
Formula
Related Topic C-Set SEE (SEP)
The SEE and SEP provide the magnitude of the standard deviation for the calibration set and the
independent validation set. The two values should be as small as possible, but they are likely to be
comparable with the standard deviation of the conventional laboratory method. With an acceptable
calibration, the two values are also roughly equal (Consistency: around 100).
Statistically the expected error is with a probability of 68 % within an interval of ± SEP and with a
probability of 95 % within an interval of ± 2 SEP.
If the SEP is significantly higher than the SEE, the calibration is overfitted (too many secondary PCs
are selected).