On sample-size and power calculations for studies using confidence intervals

Area of Science:

• Epidemiology
• Biostatistics

Background:

• Shift from significance testing to confidence intervals in epidemiologic analysis.
• Proposed use of expected confidence intervals for study design.

Purpose of the Study:

• To evaluate the utility and potential pitfalls of expected confidence intervals in epidemiologic study design.
• To propose methods for ensuring reliable discriminatory power using confidence intervals.

Main Methods:

• Analysis of expected confidence intervals in relation to study power.
• Application of conventional power and sample size formulas.
• Design strategies to ensure confidence intervals exclude incorrect parameter values.

Main Results:

• Improperly centered expected confidence intervals can be misleading indicators of study discriminatory power.
• Designing studies to ensure intervals exclude incorrect values is crucial.
• Properly centered expected intervals can enable uniformly powerful studies.

Conclusions:

• Expected confidence intervals require careful centering to accurately reflect study discriminatory power.
• Study design should prioritize ensuring intervals exclude plausible but incorrect parameter values.
• While enabling uniformly powerful studies, properly centered expected intervals may increase sample size requirements compared to other methods.