A Bayesian Approach to Account for Misclassification and Overdispersion in Count Data

Area of Science:

• Epidemiology
• Biostatistics
• Statistical Modeling

Background:

• Count data frequently contain non-sampling errors, including misclassification, measurement error, and unmeasured confounding.
• These errors can introduce significant bias into statistical estimators, compromising research findings.
• Sensitivity analyses are crucial for epidemiologists to address and quantify the impact of such errors.

Purpose of the Study:

• To extend existing Poisson regression models to effectively handle misclassification in count data.
• To incorporate extra-Poisson variability using random effects to account for overdispersion.
• To demonstrate the benefits of these enhanced models through simulation studies.

Main Methods:

• Development of Poisson regression models incorporating misclassification adjustments.
• Inclusion of random effects to model extra-Poisson variability (overdispersion).
• Simulation studies to evaluate model performance and compare inference with and without adjustments.

Main Results:

• The proposed models provide a robust framework for analyzing count data with misclassification.
• Accounting for both misclassification and overdispersion leads to substantial improvements in the accuracy of estimators.
• Simulations confirm the enhanced inferential capabilities of the adjusted models.

Conclusions:

• Epidemiologists should integrate sensitivity analyses and employ advanced statistical models to mitigate non-sampling errors.
• The extended Poisson regression models offer a valuable tool for more reliable analysis of biased count data.
• Addressing misclassification and overdispersion is essential for robust epidemiological inference.