Calibration transfer between NIR instruments using optimally predictive calibration subsets.

Area of Science:

• Analytical Chemistry
• Chemometrics
• Spectroscopy

Background:

• Calibration model transfer between Near-Infrared (NIR) instruments is crucial for consistent analysis.
• Selecting informative standardization samples is key to efficient and accurate model transfer.
• Existing methods may not always yield the most optimal or minimal set of samples.

Purpose of the Study:

• To propose and evaluate a novel approach for selecting informative standardization samples.
• To reduce the number of calibration samples required for effective NIR model transfer.
• To compare the proposed method with existing techniques like Kennard-Stone.

Main Methods:

• Development of a calibration model using Final Complexity Adapted Models (FCAM) with significance of PLS regression coefficients (FCAM-SIG).
• Selection of an optimally predictive calibration subset (OPCS) from the initial calibration set.
• Application of Piecewise Direct Standardization (PDS) for spectral transfer and optimization using response surface plots.
• Evaluation of sample set size and predictive performance (RMSEP).

Main Results:

• The proposed OPCS approach identified statistically significant smaller standardization sets compared to the Kennard-Stone method.
• Predictive performances of models using OPCS-selected samples were comparable to those using traditional methods.
• The study demonstrated the effectiveness of OPCS in optimizing sample selection for NIR model transfer.

Conclusions:

• The OPCS method offers an efficient strategy for selecting standardization samples in NIR model transfer.
• This approach leads to reduced sample sizes without compromising predictive accuracy.
• OPCS provides a valuable tool for optimizing calibration model transfer across different NIR instruments.