Precision Profile Weighted Deming Regression for Methods Comparison

Area of Science:

• Analytical Chemistry
• Biostatistics
• Method Validation

Background:

• Method validation commonly uses methods comparison (MC) with a reference method.
• Parametric modeling, specifically Deming regression, is the most robust approach for MC.
• Deming regression requires accurate weighting based on measurement variance, which is often concentration-dependent.

Purpose of the Study:

• To present the mathematical framework for integrating precision profile models with errors-in-variables (Deming) regression.
• To provide a method for weighted Deming regression applicable when precision profiles are known or unknown.
• To enhance the accuracy and generalizability of method validation studies.

Main Methods:

• The study outlines the mathematical theory connecting precision profile models with Deming regression.
• Weighted Deming regression is implemented using R code for both known and unknown precision profile scenarios.
• The implementation includes diagnostics such as jackknife standard errors, confidence intervals, and outlier testing.

Main Results:

• R code is provided for weighted Deming regression with known and unknown precision profiles.
• The methodology supports diagnostics for normality, linearity, and outlier identification.
• This approach overcomes limitations of previous methods that assumed constant variance or coefficient of variation.

Conclusions:

• Precision profile models offer a more flexible and generalizable approach to Deming regression compared to existing methods.
• This framework allows for accurate method validation even when measurement variance changes with analyte concentration.
• The developed R codes facilitate the practical application of these advanced statistical methods.