Comparison of Bayesian and frequentist methods for prevalence estimation under misclassification

Area of Science:

• Biostatistics
• Epidemiology
• Medical Diagnostics

Background:

• Statistical inference for prevalence estimation often encounters challenges with imperfect diagnostic tests, leading to misclassifications.
• Traditional methods can result in truncated prevalence estimates and confidence intervals, along with poor coverage performance.

Purpose of the Study:

• To validate a Bayesian prevalence estimation method using simulated data.
• To compare the performance of the Bayesian method against frequentist approaches, specifically the Rogan-Gladen estimate (RGE).

Main Methods:

• Utilized simulated datasets to assess estimation accuracy and confidence interval properties.
• Compared Bayesian credible intervals with frequentist confidence intervals, including the RGE and Lang-Reiczigel methods.
• Evaluated point estimate error distribution, confidence interval coverage, and interval length.

Main Results:

• Bayesian and RGE point estimates showed similar error distributions, with a slight advantage for the Bayesian method.
• The Bayesian estimate avoided the truncation issues observed with the RGE.
• Frequentist confidence intervals demonstrated significant under-coverage, while Bayesian credible intervals and the Lang-Reiczigel method performed well.

Conclusions:

• The Bayesian prevalence estimation method is recommended over traditional frequentist methods for improved accuracy and reliability.
• Combining the Rogan-Gladen point estimate with the Lang-Reiczigel confidence interval is a viable alternative.