Uniformity test of bias when the reference value contains experimental error

Area of Science:

• Analytical Chemistry
• Method Validation
• Statistical Analysis

Background:

• Uniformity tests for analytical method biases must account for reference method uncertainties.
• Ignoring significant uncertainty in true value estimates can lead to Type I errors exceeding specified rates.
• Confounding bias, a random component mixed with fixed bias, complicates standard bias equality tests.

Purpose of the Study:

• To develop a statistical test that accurately assesses the uniformity of fixed biases in analytical methods.
• To address the issue of inflated Type I error rates caused by confounding bias and reference method uncertainty.
• To propose a method that eliminates the impact of confounding bias on fixed bias uniformity testing.

Main Methods:

• Development of a confounding model incorporating both fixed and confounding biases.
• Calculation of the actual Type I error rate for uniformity tests under the confounding model.
• Introduction of a novel F -test, which modifies the conventional F-test with a correction factor.
• The correction factor is derived from the uncertainty of true value estimates.

Main Results:

• The proposed F -test effectively reduces the Type I error rate to the prespecified level, as confirmed by simulation studies.
• The F -test demonstrated practical application using data from two aldehyde analytical methods.
• The study provides recommendations for optimal sample allocation in bias uniformity testing.

Conclusions:

• The F -test provides a statistically sound approach to uniformity testing of fixed biases by accounting for reference method uncertainty.
• Accurate bias uniformity testing is crucial for reliable analytical method validation.
• The proposed method enhances the reliability of analytical method performance evaluation.