Marginal Bayesian Semiparametric Modeling of Mismeasured Multivariate Interval-Censored Data

Area of Science:

• Biostatistics
• Statistical Modeling
• Oral Health Research

Background:

• Correlated time-to-event data are common in longitudinal studies, such as oral health research.
• Challenges include interval-censored responses and potential misclassification of event occurrences.
• Existing models often struggle to account for both complexities simultaneously.

Purpose of the Study:

• To propose a novel Bayesian nonparametric approach for population-averaged modeling of correlated time-to-event data.
• To address challenges of interval censoring and misclassification in event determination.
• To compare different semiparametric models including proportional hazards, proportional odds, and accelerated failure time.

Main Methods:

• Development of a joint semiparametric model for unobserved time-to-event data.
• Utilizing a flexible tailfree prior for the baseline distribution.
• Incorporation of a parametric copula function and a detailed misclassification model, accounting for multiple examiners.

Main Results:

• Empirical evidence demonstrates the model's ability to estimate time-to-event distributions and misclassification parameters without external information.
• The study quantifies the impact of ignoring misclassification on statistical inferences.
• Successful application in an oral health study context.

Conclusions:

• The proposed Bayesian nonparametric approach provides a robust framework for analyzing complex time-to-event data.
• Accurate modeling of interval censoring and misclassification is essential for reliable statistical inference.
• The methodology offers valuable insights for oral health research and other fields with similar data characteristics.