JAGS model specification for spatiotemporal epidemiological modelling.

Area of Science:

• Epidemiology
• Computational Statistics
• Infectious Disease Dynamics

Background:

• Bayesian inference using Gibbs Sampling (BUGS) has become prominent in infectious disease modeling over the last two decades.
• The integration of Markov chain Monte Carlo (MCMC) methods has popularized Bayesian analysis in this field.
• Complex infectious disease models with spatial-temporal components and numerous parameters present computational challenges for existing MCMC software.

Purpose of the Study:

• To investigate and compare the performance of two alternative subscripting strategies for model creation within the Just Another Gibbs Sampler (JAGS) environment.
• To assess the impact of these strategies on computational run times for Bayesian spatiotemporal infectious disease models.

Main Methods:

• Implementation of two distinct subscripting strategies for defining models in JAGS.
• Empirical evaluation of the run times associated with each strategy using complex infectious disease models.
• Focus on models incorporating spatial and temporal dependencies and multiple parameters.

Main Results:

• The study identified significant differences in run times between the two investigated subscripting strategies.
• One strategy demonstrated superior performance in terms of computational efficiency for the tested models.
• The findings provide practical insights into optimizing model implementation in JAGS.

Conclusions:

• The choice of subscripting strategy in JAGS can substantially impact the efficiency of Bayesian spatiotemporal infectious disease modeling.
• Practitioners can leverage these findings to select more efficient modeling approaches, ensuring timely analysis.
• This research contributes to the practical application of advanced computational techniques in epidemiological studies.