State-space modelling for infectious disease surveillance data: Stochastic simulation techniques and structural...

Area of Science:

• Computational statistics
• Epidemiology
• Mathematical modeling

Background:

• Infectious disease modeling requires robust statistical techniques to capture dynamic transmission patterns.
• State-space models offer a flexible framework for analyzing time-series data in public health.

Purpose of the Study:

• To explore advanced stochastic simulation techniques for state-space models in infectious disease analysis.
• To apply these methods to COVID-19 surveillance data for detecting structural changes and forecasting trends.
• To evaluate the impact of different prior distributions on model performance.

Main Methods:

• Utilized Markov Chain Monte Carlo (MCMC) and Sequential Monte Carlo (SMC) methods.
• Applied Kalman smoothing and Bayesian inference for non-linear dynamic regression models.
• Assessed various priors (normal, Student's t, Laplace, horseshoe) with a Rao-Blackwellized particle filter.

Main Results:

• The horseshoe prior exhibited superior performance in identifying change points and adapting to complex data structures.
• Stochastic simulation techniques effectively detected structural changes in COVID-19 case counts.
• Models provided insights for real-time monitoring and forecasting.

Conclusions:

• State-space models, enhanced with sophisticated prior distributions like the horseshoe prior, offer a nuanced understanding of infectious disease transmission.
• Advanced simulation techniques are crucial for accurate public health surveillance and prediction.
• The findings support the use of these methods for real-time infectious disease management.