Bayesian non-parametric inference for stochastic epidemic models using Gaussian Processes.

Area of Science:

• Epidemiology
• Biostatistics
• Computational Biology

Background:

• Traditional epidemic models often rely on restrictive assumptions about disease transmission rates.
• These assumptions are rarely tested, potentially limiting model accuracy and real-world applicability.
• There is a need for more flexible statistical methods to analyze epidemic data.

Purpose of the Study:

• To develop and evaluate novel Bayesian non-parametric methods for stochastic epidemic models.
• To relax restrictive assumptions inherent in standard epidemic modeling techniques.
• To accurately estimate the underlying infection process using flexible statistical approaches.

Main Methods:

• Development of Bayesian non-parametric models utilizing Gaussian Processes.
• Application of these methods to estimate the infection process in epidemic models.
• Validation using both simulated and real-world epidemiological datasets.

Main Results:

• Simulated data demonstrated that the proposed methods can effectively recover the true infection process.
• Real-world data analysis showed successful application of the methods across different epidemiological settings.
• The Gaussian Process approach provides a more flexible alternative to traditional modeling assumptions.

Conclusions:

• Bayesian non-parametric methods, particularly those using Gaussian Processes, offer a powerful and flexible framework for stochastic epidemic modeling.
• These methods relax rigid assumptions, leading to more accurate estimation of infection dynamics.
• The approach is validated and applicable to diverse epidemiological scenarios, enhancing disease surveillance and control strategies.