Inferences on the within-subject coefficient of variation

Area of Science:

• Biostatistics
• Medical Statistics
• Biological Data Analysis

Background:

• The within-subject coefficient of variation (WCV) is a standard metric for assessing data precision and reproducibility in biological and medical sciences.
• Accurate statistical methods are crucial for reliable WCV estimation, particularly in scenarios with limited data.

Purpose of the Study:

• To develop novel statistical procedures for constructing generalized confidence intervals and hypothesis tests for a single within-subject coefficient of variation (WCV).
• To extend these methods for comparing WCV between two samples.
• To provide statistically sound and computationally efficient tools for WCV analysis.

Main Methods:

• Utilized the concept of generalized pivots under a one-way random effect model to derive procedures for a single WCV.
• Extended the generalized pivot approach to address two-sample WCV comparisons.
• Focused on methods ensuring good coverage probabilities and type-I error control.

Main Results:

• Developed generalized confidence intervals and tests for single and two-sample WCV.
• Demonstrated that the proposed methods are easy to compute.
• Showcased favorable performance in terms of coverage probabilities and type-I error control, particularly with small sample sizes.

Conclusions:

• The proposed generalized pivot-based methods provide a robust framework for WCV analysis in biostatistics.
• These methods are effective and reliable, especially for small sample sizes common in biological and medical studies.
• The study offers practical statistical tools illustrated with a real-world example.