Table 3.
Prediction perfomances of the N and C models.
| n | n outliers | n CAL (n fix / n rotating) | n VAL (n fix / n rotating) | PLS-R factors | Spectra pretreatment | Concentration range (%DM) | Spectral range (wave numbers in cm−1) | RMSEC | RMSEP | RPD | R² | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| N model | 2427 | 15 | 1205 (1116/89) | 1207 (1114/93) | 17 | SNV | 0.2–5.9 | 4242.9–7506.1 | 0.244 | 0.248 | 5.93 | 0.97 |
| C model | 2427 | 8 | 1210 (1118/92) | 1209 (1110/99) | 7 | SNV and first derivate | 21.5–47.5 | 6094.4–9403.8 | 1.31 | 1.32 | 2.12 | 0.78 |
n, number of samples; CAL, calibration; VAL, validation; fix, fixed glass cup of 20 mm used for NIR spectra acquisition; rotation, rotative quartz cup of 51 mm ∅ used for NIR spectra acquisition; PLS-R, partial least squares regression; RMSEC : root mean square error in calibration; RMSEP : root mean square error in prediction; RPD : ratio of performance deviation; R² : coefficient of determination.