Combinatorial Regression for Analytical Chemistry.

Area of Science:

• Analytical Chemistry
• Spectrometry

Background:

• Current analytical chemistry guidelines for method validation may weaken statistical validity by using few standards and no blanks.
• Excellent precision in measurements does not guarantee good accuracy, a limitation highlighted by the pooled calibration (PoPC) paradigm.

Purpose of the Study:

• To introduce combinatorial regression (CR) as a novel technique to resolve disagreements in computing measurement uncertainty from calibration lines.
• To evaluate the effectiveness of CR in estimating slopes, intercepts, and standard deviations for improved uncertainty assessment.

Main Methods:

• Development and application of combinatorial regression (CR) for data analysis.
• Evaluation of CR using high-resolution continuous-source flame atomic absorption spectrometry (HR-CR FAAS) for copper determination.
• Comparison of CR results with existing IUPAC equations.

Main Results:

• CR provides estimates for slopes, intercepts, and standard deviations, addressing limitations of current methods.
• Standard deviations of replicates were found to be too large for direct uncertainty application under current models.
• Concentration residuals from CR emerged as viable estimates for measurement uncertainty, unlike the small standard deviations from IUPAC equations.
• CR yielded coefficients of variation (CVs) that increased with concentration and replicate number, suggesting its suitability for low-concentration analyses.

Conclusions:

• Combinatorial regression (CR) offers a promising approach to enhance the accuracy of measurement uncertainty calculations in analytical chemistry.
• CR demonstrates advantages over traditional IUPAC equations, particularly for low-concentration samples analyzed via HR-CR FAAS.
• The study highlights the importance of addressing both precision and accuracy in method validation and uncertainty estimation.