Modeling and inference for infectious disease dynamics: a likelihood-based approach.

Area of Science:

• Epidemiology
• Statistical Modeling
• Computational Biology

Background:

• Likelihood-based statistical inference is widely used in stochastic modeling across scientific fields.
• Mechanistic models in infectious disease dynamics capture biological processes but often lack closed-form likelihood expressions, leading to computational challenges.

Purpose of the Study:

• To review novel data-motivated mechanistic models developed over the last decade.
• To highlight key statistical aspects, such as overdispersion, in the interface of nonlinear infectious disease modeling and data analysis.
• To suggest future directions for mechanistic model exploration.

Main Methods:

• Review of recent literature on mechanistic models in infectious disease dynamics.
• Focus on statistical inference methods, particularly likelihood maximization.
• Discussion of computational algorithms that facilitate likelihood maximization for models lacking closed-form expressions.

Main Results:

• Algorithmic advancements have significantly eased likelihood maximization for complex mechanistic models.
• This has spurred the development and study of novel data-motivated mechanistic models.
• Overdispersion is identified as a critical statistical factor in nonlinear infectious disease modeling and data analysis.

Conclusions:

• Recent algorithmic progress has overcome computational hurdles in likelihood-based inference for mechanistic disease models.
• This enables the use of more sophisticated, data-driven mechanistic models.
• Further research into statistical aspects like overdispersion is essential for advancing infectious disease modeling.