Table 2.

Comparison of accuracies (mean ± SD) achieved for full spectrum PLSR versus best model from variable selection using CARS-PLS resulting from 100 repetitions for the calibration and validation.

Models (S = 100)	WL-Selection	Full spectrum PLSR	CARS-PLSR
Calibration	Variables	662	16
	nLV	7.37 ± 1.28	6.99 ± 0.58
	RMSE [%]	3.58 ± 0.54	2.75 ± 0.30
Validation	RPD	1.96 ± 0.35	2.67 ± 0.46
	RP2	0.74 ± 0.08	0.86 ± 0.05
	RMSEP [%]	3.83 ± 0.57	2.82 ± 0.40
	SEP [%]	3.79 ± 0.56	2.78 ± 0.40
	BIAS	−0.22 ± 0.86	−0.19 ± 0.65

Abbreviations are as given in Table 1. In addition we use root mean square error of calibration (RMSE).