An augmented classical least squares method for quantitative Raman spectral analysis against component information...

Area of Science:

• Analytical Chemistry
• Spectroscopy

Background:

• Quantitative Raman spectral analysis often faces challenges due to component information loss.
• Classical Least Squares (CLS) calibration can be sensitive to incomplete spectral information.

Purpose of the Study:

• To introduce an Augmented Classical Least Squares (ACLS) calibration method.
• To improve the robustness of quantitative Raman spectral analysis against component information loss.

Main Methods:

• Selected Raman spectral signals with low analyte concentration correlations as substitutes for unknown components.
• Determined the optimal number of selected signals using leave-one-out root-mean-square error of cross-validation (RMSECV).
• Constructed an ACLS model using an augmented concentration matrix and reference spectral data.

Main Results:

• The ACLS method demonstrated increased robust predictive power compared to traditional CLS.
• Evaluated predictive power using 2-fold cross-validation and Venetian blinds strategy.
• One-way variance analysis (ANOVA) confirmed comparable predictive power to Partial Least Squares (PLS) and Principal Component Regression (PCR).

Conclusions:

• The proposed ACLS method effectively addresses component information loss in quantitative Raman spectroscopy.
• ACLS offers a robust alternative for spectral calibration, achieving performance similar to PLS and PCR.