Table 3.
Summary of cross-validation statistics of mPLS models used for the prediction of biomass quality traits from NIRS spectral data.
| Constituent | Samples* | Chemical analysis | NIRS prediction | r2¥ | SECV§ | ||||
|---|---|---|---|---|---|---|---|---|---|
| Mean | Min | Max | Mean | Min | Max | ||||
| NDF (g/kg dm) | 246 | 85.04 | 71.55 | 92.69 | 85.04 | 71.28 | 92.35 | 0.99 | 0.88 |
| ADF (g/kg dm) | 243 | 54.96 | 38.43 | 68.55 | 54.97 | 39.40 | 68.47 | 0.99 | 1.13 |
| ADL (g/kg dm) | 239 | 9.22 | 4.88 | 14.45 | 9.20 | 5.26 | 14.42 | 0.88 | 0.79 |
| Cellulose conversion (%) | 237 | 29.89 | 8.17 | 52.10 | 30.21 | 13.14 | 46.81 | 0.92 | 3.22 |
| Hemicellulose conversion (%) | 243 | 12.43 | 5.84 | 22.20 | 12.34 | 6.70 | 20.27 | 0.82 | 2.06 |
*
, Sample number varies as for every trait different samples may be removed by the software as outliers; depending on the model
¥
r2, coefficient of determination;
§
SECV, Standard error of cross-validation.