The relationship among design parameters for statistical power between continuous and binomial outcomes in cluster...

Area of Science:

• Statistics
• Biostatistics
• Research Methodology

Background:

• Extensive research exists on statistical power for continuous outcomes.
• Literature on statistical power for binary outcomes is comparatively limited.
• A key challenge is establishing analogies between continuous and binary design parameters.

Purpose of the Study:

• To investigate methods for estimating statistical power for binary outcomes.
• To explore analogies for the intraclass correlation coefficient (ICC) in binary data.
• To assess the implications of lacking a useful ICC analogy for statistical inference.

Main Methods:

• Proposed two potential analogies for the ICC in binary outcome designs.
• Conducted two simulation studies to compare power estimates and Type I error rates.
• Analyzed results under two analytic models for binary outcomes and ICC analogies.

Main Results:

• Identified challenges in establishing valid and useful ICC analogies for binary outcomes.
• Compared power estimates and Type I error rates between analytic models and ICC analogies.
• Assessed the impact on statistical inference when no direct ICC analogy exists.

Conclusions:

• Acknowledged limitations in creating direct analogies for design parameters between continuous and binary outcomes.
• Provided an empirical example to guide researchers in conceptualizing analogies.
• Suggested directions for future research on statistical power for binary data.