Approximate Bayesian inference for joint linear and partially linear modeling of longitudinal zero-inflated count and...

Area of Science:

• Biostatistics
• Statistical Modeling
• Longitudinal Data Analysis

Background:

• Joint modeling of count and survival data is crucial in various fields.
• Traditional methods often rely on computationally intensive Markov Chain Monte Carlo (MCMC).
• Latent Gaussian models provide a framework for such complex data structures.

Purpose of the Study:

• To propose an efficient approximate Bayesian approach for joint modeling of zero-inflated count and time-to-event data.
• To introduce a joint partially linear model to capture non-linear time effects in longitudinal count responses.
• To compare the performance of the proposed method against MCMC using simulations and real-world data.

Main Methods:

• Utilized Integrated Nested Laplace Approximation (INLA) for approximate Bayesian inference.
• Developed a zero-inflated hurdle model (Poisson or negative binomial) for count data.
• Employed a Weibull model for survival time data.
• Considered both linear and partially linear joint models.

Main Results:

• The INLA approach demonstrated good performance in simulation studies.
• The proposed method showed comparable or superior efficiency to MCMC.
• Successful application to pregnancy and HIV study datasets.

Conclusions:

• INLA provides an efficient computational tool for joint modeling of zero-inflated count and time-to-event data.
• The proposed joint partially linear model effectively handles non-linear time effects.
• The method is applicable to real-world biomedical research.