Quantifying the Prediction Error in Analytical Multivariate Curve Resolution Studies of Multicomponent Systems

Area of Science:

• Analytical Chemistry
• Chemometrics

Background:

• Multivariate curve resolution (MCR) analysis often results in ambiguous solutions due to rotational ambiguities in bilinear decomposition.
• Quantitative analysis in MCR, especially with uncalibrated interferents, faces challenges in univocally recovering analyte profiles in test samples.

Purpose of the Study:

• To introduce a quantitative measure for prediction errors arising from rotational ambiguities in MCR.
• To provide a method for estimating absolute and relative quantitative errors caused by rotational ambiguities.

Main Methods:

• Utilized the MCR-BANDS procedure to calculate concentration profiles.
• Quantified rotational ambiguity errors by measuring differences in the area under analyte concentration profiles.
• Applied the methodology to simulated and experimental data from liquid chromatography with diode array detection.

Main Results:

• Developed a quantitative measure to assess prediction errors stemming from MCR rotational ambiguities.
• Demonstrated the estimation of both absolute and relative quantitative errors.
• Validated the methodology across diverse analytical scenarios with multiple analytes and interferents.

Conclusions:

• The proposed method effectively quantifies prediction errors due to rotational ambiguities in MCR analysis.
• This approach enhances the reliability of quantitative MCR applications, particularly in complex matrices.
• The methodology is generalizable to various analytical techniques generating data tables or matrices.