Table 1.
Regression parameters from PLS models for the three NIR techniques. The three models have different complexities (number of latent variables, LVs), and the parameters used for describing the performances are the coefficient of determination (R2) and the root mean squared errors (RMSEs). The subscripts stand for CAL = calibration, CV = cross validation, and PRED = prediction (test set).
| SCiO | MPA Sphere | MPA Probe | |
|---|---|---|---|
| LVs | 7 | 4 | 3 |
| R2CAL | 0.717 | 0.925 | 0.846 |
| R2CV | 0.461 | 0.903 | 0.793 |
| R2PRED | 0.550 | 0.713 | 0.897 |
| RMSECAL | 0.966 | 0.566 | 0.870 |
| RMSECV | 1.332 | 0.645 | 1.008 |
| RMSEPRED | 1.217 | 1.109 | 0.712 |
| RPDCAL | 1.904 | 3.709 | 2.583 |
| RPDCV | 1.380 | 3.254 | 2.229 |
| RPDPRED | 1.909 | 1.773 | 2.167 |