[Comparison of 7 methods for sample size determination based on confidence interval estimation for a single...

Area of Science:

• Statistics
• Biostatistics
• Epidemiology

Context:

• Accurate sample size calculation is fundamental for reliable statistical inference.
• Confidence intervals are widely used to estimate population parameters, such as proportions.
• The choice of sample size method can significantly impact study precision and power.

Purpose:

• To compare the performance of seven distinct sample size calculation methods for single proportions.
• To evaluate these methods across various event incidences and desired precision levels (confidence interval widths).
• To identify optimal methods based on simulation results for width, coverage, and noncoverage probability.

Summary:

• Seven methods (Wald, Agresti-Coull add z2, Agresti-Coull add 4, Wilson Score, Clopper-Pearson, Mid-p, Jefferys) were assessed using Monte Carlo simulations.
• For high precision (ω=0.05), Mid-p and Clopper-Pearson showed superior performance with low event incidence (P < 0.15).
• The Clopper-Pearson method demonstrated robustness, particularly in low event incidence settings (P=0.01-0.05) where other methods encountered iteration failures.

Impact:

• Provides evidence-based guidance for selecting appropriate sample size calculation methods in proportion studies.
• Highlights the importance of considering event incidence and precision requirements when determining sample size.
• Contributes to improving the rigor and efficiency of study designs in various scientific fields.