Semi-Bayes and empirical Bayes adjustment methods for multiple comparisons

Area of Science:

• Epidemiology
• Biostatistics
• Occupational Health

Background:

• Epidemiological studies frequently involve numerous statistical tests, increasing the risk of false positive findings.
• Traditional multiple comparison adjustments, like the Bonferroni method, can be overly conservative and may lead to the dismissal of important results.
• The Bonferroni method adjusts p-values but not effect estimates (e.g., odds ratios), potentially limiting the validity of findings.

Purpose of the Study:

• To evaluate the performance of Empirical Bayes and semi-Bayes methods in addressing false positives in epidemiological research.
• To compare the efficacy of these Bayesian approaches against traditional methods in a real-world study.
• To assess the impact of these methods on the validity of effect estimates.

Main Methods:

• Application of Empirical Bayes and semi-Bayes statistical methods.
• Utilized a case-control study design focusing on occupational risk factors for lung cancer.
• Performance testing of the applied Bayesian methods.

Main Results:

• Empirical Bayes and semi-Bayes methods demonstrated an ability to mitigate numerous false positive associations.
• These methods yielded, on average, more valid effect estimates compared to traditional approaches.
• The study successfully tested the performance of these advanced statistical techniques.

Conclusions:

• Empirical Bayes and semi-Bayes methods offer a superior approach to handling multiple comparisons in epidemiological studies.
• These methods enhance the reliability of statistical findings and the accuracy of effect size estimations.
• The application in a lung cancer occupational risk study validates their utility in public health research.