Forecasting models

Nowcasting and forecasting of infectious disease dynamics

• We can classify models by the level of mechanism they include
• All of the model types we will introduce in the next few slides have been used for COVID-19 forecasting (the US and/or European COVID-19 forecast hub)

NOTE: level of mechanism \(\neq\) model complexity

Cramer et al., Scientific Data, 2022

Complex agent-based models

Conceptually probably the closest to metereological forecasting, if with much less real-time data.

Venkatramanan et al., Epidemics, 2018

Compartmental models

Aim to capture relevant mechanisms but without going to the individual level.

Keeling et al., PLOS Comp Biol, 2021

Semi-mechanistic models

• Include some epidemiological mechanism (e.g. SIR or the renewal equation)
• Add a nonmechanistic time component inspired by statistical models (e.g. random walk)
• e.g., the model we have worked with so far

Dureau et al., Biostatistics, 2013

Statistical models

• Models that don’t include any epidemiological background e.g. ARIMA; also called time-series models
• The random walk model when used on its own (without going via \(R_t\)) is called a stochastic volatility model

Other models

• Expert or crowd opinion
• Machine learning
• …

Bosse et al., Wellcome Open Res, 2024

Your Turn

1. Review the performance of the random walk model
2. Motivate a mechanism to order to address some of the issues
3. Motivate a statistical approach aiming to address the same issues
4. Compare the performance of these models for a single forecast
5. Evaluate many forecast from these models and compare their performance

Return to the session