Table 4. Summary of statistical analyses of TNCs-DC in calibration and prediction sets.

Samples	Calibration set				Prediction set
	N.[1]	Max.[3] (%)	Min.[4] (%)	Mean (%) ± S.D.[2]	N.[1]	Max.[3] (%)	Min.[4] (%)	Mean (%) ± S.D.[2]
Leaf	90	4.871	2.264	3.357±0.701	30	4.731	2.398	3.108±0.612
Root	30	1.470	0.847	1.183±0.156	10	1.301	1.035	1.194±0.103
Stem	90	2.135	0.556	1.040±0.403	30	1.636	0.599	0.849±0.218
All	210	4.871	0.556	2.114±1.292	70	4.731	0.599	1.927±1.218

Note:

^[1]N.: Number of samples;

^[2]S. D.: Standard deviation of the group;

^[3]Max.: Maximum;

^[4]Min.: Minimum;

Calibrations set with wide variation range of TNCs-DC could benefit for building robust models. Cross-validation set had the same results with the calibration set, which were not motioned in this table. In this study, leave-one-out cross-validation (one sample randomly chosen from calibration set was retained at a time and the rest of samples in calibration set were used to build the model) was used to verify the reproducibility and robustness of models.