Assessing the Impact of Misclassification in Case-Control Studies: An Examination of Bayesian Credible Intervals 公开

Abstract

Objectives: Case-control studies are fundamental in identifying associations between exposures and rare diseases, yet the accuracy of odds ratio (OR) estimates can be compromised by misclassification. Traditional Wald confidence intervals have limitations, especially with small sample sizes or extreme proportions. Bayesian methods utilizing Jeffreys priors have been proposed as a robust alternative, yet empirical evidence on their comparative efficacy is sparse when misclassification is an issue. This study aimed to evaluate the performance of Bayesian methods with Jeffreys priors against traditional Wald approaches in interval estimation for the exposure odds ratio, focusing on interval width and frequentist coverage in the context of case- control studies with main and internal validation data.

Methods: We analyzed a real dataset to compare a naïve log odds ratio estimate versus an adjusted (via maximum likelihood) estimate, and to compare the Wald-type confidence interval with a proposed credible interval based on a Jeffreys Dirichlet prior. Then, we conducted simulations with smaller case and control sample sizes to compare the performance of Wald-based and Bayesian credible intervals in terms of interval width and coverage probability. The simulations were designed to reflect a realistic range of exposure misclassification scenarios encountered in epidemiological research.

Results: Our simulations demonstrated that the proposed Bayesian credible interval, compared to the Wald interval, offered significantly narrower interval widths while maintaining near-nominal frequentist coverage across a variety of exposure odds ratios and misclassification scenarios. Specifically, Bayesian methods provided more precise interval estimates and favorable coverage probabilities, underlining their potential for more accurate and reliable statistical inference in case-control studies.

Conclusions: The findings suggest that Bayesian methods utilizing Jeffreys priors represent a significant advancement over traditional Wald intervals for the estimation of exposure odds ratios in case-control studies featuring main and internal validation study data. This study supports the adoption of Bayesian approaches in epidemiological research, especially in the presence of misclassification and when precise interval estimation is critical.

Keywords: Case-control studies, Bayesian methods, Jeffreys priors, Misclassification. 

Table of Contents

Introduction

Methods

Results

Discussion

About this Master's Thesis

Rights statement

• Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.

School	Rollins School of Public Health
Department	Biostatistics
Subfield / Discipline	Biostatistics - MPH & MSPH
Degree	M.S.P.H.
Submission	Master's Thesis
Language	English
Research Field	Epistemology Statistics
关键词	Bayesian methods Case-control studies Jeffreys priors Misclassification
Committee Chair / Thesis Advisor	Robert Lyles, Emory University

最新修改

Primary PDF

Thumbnail	Title	Date Uploaded	Actions
	Assessing the Impact of Misclassification in Case-Control Studies: An Examination of Bayesian Credible Intervals ()	2024-04-08 09:46:41 -0400	Press to Select an action Download

Supplemental Files

Thumbnail	Title	Date Uploaded	Actions