An Augmented Classical Least Squares Method for Quantitative Raman Spectral Analysis against Component Information Loss

Table 1

Prediction power comparison.

Method	Validation set			Calibration set
	𝑅^2	RMSEP	𝑃 value	RMSECV
CLS1	0.02818	0.24022	—	0.244141
CRACLS	0.99423	0.01275	<0.0001	0.01573
CLS2	0.99429	0.01232	0.5483	0.01564
PCR	0.99417	0.01212	0.1149	0.01412
PLS	0.99467	0.01165	0.3467	0.01318
Proposed method	0.99534	0.01128	0.2568	0.01303

The models are sorted according to increasing prediction power, and the 𝑃 values for the significance testing by a one-way ANOVA of the improvement compared to the previous model are given.
CLS1: classical least squares using only the analyte concentration of interesting.
CLS2: classical least squares using all analyte concentrations.
PCR: principal component regression.
PLS: partial least squares.
CRACLS: concentration residual augmented classical least squares.
RMSECV: root-mean-square error of cross-validation.
RMSEP: root-mean-square error of prediction.