Power and sample size for reversible linear mixed models with clustering and longitudinality: GLIMMPSE Version 3

Area of Science:

• Biostatistics
• Statistical Software Development
• Clinical Trial Design

Background:

• General linear mixed models (GLMMs) are crucial for analyzing complex data structures in research.
• Accurate power and sample size calculations are essential for efficient study design and resource allocation.
• Existing software may have limitations in handling multilevel, longitudinal, or combined data structures.

Purpose of the Study:

• To introduce GLIMMPSE Version 3, an updated, free, web-based, open-source software tool.
• To enhance the calculation of power and sample size for general linear mixed models with Gaussian errors.
• To provide a user-friendly interface and advanced features for complex study designs.

Main Methods:

• Refactored back end in Python for improved performance.
• Simplified user interface with single-topic screens for ease of use.
• Implemented a recursive algorithm for computing covariances up to ten levels of clustering.
• Utilized updated Monte Carlo simulations for accuracy validation.

Main Results:

• GLIMMPSE Version 3 offers power calculations for studies with clustering, repeated measures, or both.
• The software supports a wide range of testable hypotheses within various study designs.
• Updated simulations demonstrate power approximation accuracy within 0.01.
• Five new examples showcase applications in cluster-randomized trials, longitudinal, multilevel, and complex studies.

Conclusions:

• GLIMMPSE Version 3 provides a robust and accessible tool for power and sample size calculations in complex research.
• The software's new features and improved accuracy support the design of efficient and statistically sound studies.
• Its widespread use and NIH funding underscore its importance in biomedical and experimental research.