Using the coefficient of correlation in method-comparison studies.

Area of Science:

• Biostatistics
• Analytical Chemistry
• Clinical Laboratory Science

Background:

• The coefficient of correlation (R) is frequently used in method-comparison studies.
• Interpretation of R is often limited, leading to potential misinterpretations.

Purpose of the Study:

• To demonstrate how the coefficient of correlation (R) can detect interference, nonlinearity, and misuse of imprecision in method-comparison studies.
• To establish a method for predicting expected R values based on method imprecisions before comparison.

Main Methods:

• Developing a predictive model for R based on the imprecisions of the compared methods.
• Implementing a statistical test to identify deviations from predicted R at the P = 0.05 significance level.
• Evaluating the test's performance using computer simulations and real-world examples.

Main Results:

• A method to predict the coefficient of correlation (R) based on inherent method imprecisions was developed.
• Discrepancies between predicted and actual R values were shown to indicate interference, nonlinearity, or imprecision misuse.
• A statistical test was validated for detecting these issues, demonstrating effectiveness through simulation.

Conclusions:

• The coefficient of correlation (R) offers more diagnostic power in method comparisons than commonly recognized.
• Predicting R based on imprecision provides a robust approach to identify systematic errors and method limitations.
• This approach enhances the reliability and interpretability of method-comparison studies in various scientific fields.