Estimating prevalence: a confidence game.

Area of Science:

• Parasitology
• Biostatistics
• Epidemiological Methods

Background:

• Prevalence estimates in parasitology often lack appropriate measures of error.
• A minimum sample size of 15 is commonly recommended but may provide a false sense of confidence.

Purpose of the Study:

• To highlight the inadequacy of simple sample size recommendations for prevalence estimates.
• To emphasize the need for appropriate statistical methods to report confidence intervals accurately.

Main Methods:

• Analysis of the nonlinear relationship between sample size and standard error.
• Evaluation of factors influencing 95% confidence intervals for proportions.
• Consideration of Monte Carlo simulations and Bayesian posterior integration for true confidence intervals.

Main Results:

• Sample size and the estimated proportion both influence the reliability of 95% confidence intervals.
• A fixed "gold standard" sample size for reliable binomial proportion estimates does not exist.
• Commonly recommended sample sizes may not yield accurate confidence intervals.

Conclusions:

• Reporting conservative (e.g., Clopper-Pearson) or robust (e.g., Wilson score) confidence intervals is necessary.
• Accurate error reporting for prevalence estimates requires advanced statistical approaches.
• The limitations of traditional sample size recommendations in parasitological studies are evident.