A simulation study comparing tests for the equality of coefficients of variation

Area of Science:

• Biostatistics
• Medical Laboratory Science

Background:

• The coefficient of variation (CV) is a vital metric in medical and biological sciences for comparing relative variability.
• Assessing the equality of CVs across multiple populations is crucial for robust data analysis.

Purpose of the Study:

• To review and compare parametric and non-parametric statistical tests for the equality of coefficients of variation in kappa populations.
• To evaluate the performance of these tests under different distributional assumptions.

Main Methods:

• Review of existing parametric and non-parametric statistical tests.
• Conducting simulation studies to assess test performance (size and power).
• Application of tests to real-world data from a Haematology and Serology Quality Assurance Program.

Main Results:

• Parametric tests demonstrate good performance with normally distributed data but are unreliable with non-normal data.
• Non-parametric tests exhibit robustness, performing well across various underlying data distributions.
• The non-parametric test showed insensitivity to outliers in practical application.

Conclusions:

• Non-parametric tests are recommended for comparing coefficients of variation when the underlying data distribution is unknown or non-normal.
• The robustness of non-parametric methods makes them suitable for diverse biological and medical datasets, including those with potential outliers.