Using Near-infrared reflectance spectroscopy (NIRS) to predict glucobrassicin concentrations in cabbage and brussels sprout leaf tissue

Table 4 Model statistics for predicting GBS concentration on a fresh weight and dry weight bases

Spectral data used for model development	Calibration (n = 68)		Cross-validation (n = 24)
Spectral data used for model development	R²_cal^a	RMSEC^b	R²_CV^c	RMSEP^d	RPD^e	Terms^f
Raw (fr wt)	0.89	40.47	0.75	35.12	2.3	10
Raw (dr wt)	0.90	2.17	0.80	2.07	2.4	10
Raw plus SNV + DT (fr wt)	0.65	53.61	0.60	39.56	2.0	2
Raw plus SNV + DT (dr wt)	0.63	3.67	0.41	3.46	1.4	2
1^st derivative (fr wt)	0.89	43.29	0.76	30.44	2.6	6
1^st derivative (dr wt)	0.90	2.41	0.79	2.07	2.4	6
1^st derivative plus SNV + DT^g (fr wt)	0.76	53.10	0.55	42.17	1.9	2
1^st derivative plus SNV + DT (dr wt)	0.64	3.22	0.46	3.33	1.5	1

^aCoefficient of determination of the calibration
^bRoot mean squared error of calibration
^cCoefficient of determination of the cross-validation
^dRoot mean squared error of prediction
^eRatio of prediction to deviation
^fNumber of terms (PLS components) used in the model selected for cross-validation
^gStandard normal variate with detrending spectral preprocessing

ISSN: 1746-4811