tutorials

Simulation-based inference for epidemiological dynamics with Pypomp

Aaron A. King, Edward L. Ionides and Kunyang He

Overview

This short course introduces statistical inference techniques and computational methods for dynamic models of epidemiological systems. The course explores deterministic and stochastic formulations of epidemiological dynamics and develop inference methods appropriate for a range of models. Special emphasis will be on exact and approximate likelihood as the key elements in parameter estimation, hypothesis testing, and model selection. Specifically, the course emphasizes sequential Monte Carlo techniques.

This is a Python-Pypomp version of an R-pomp short course. The Pypomp library supports GPU computation and automatic differentiation via JAX.

  1. To introduce partially observed Markov process (POMP) models as tools for scientific investigation and public health policy.
  2. To give students the ability to formulate POMP models of their own.
  3. To teach efficient approaches for performing scientific inference using POMP models.
  4. To familiarize students with the Pypomp library.
  5. To help students familiar with the R-pomp package to transfer that expertise to Pypomp.
  6. To give students opportunities to work with inference methods, including methodologies that use automatic differentiation and GPUs which are not accessible via R-pomp.
  7. To provide documented examples for student re-use.
  8. All source code is available at https://github.com/pypomp/tutorials/tree/main/sbied.

Lessons

0. Instructions for preparing your laptop for the course exercises.

1. Introduction: What is “Simulation-based Inference for Epidemiological Dynamics”? POMPs and pypomp.

2. Simulation of stochastic dynamic models.

3. Likelihood for POMPs: theory and practice.

4. Iterated filtering: theory and practice.

5. Case study: measles. Recurrent epidemics, long time series, covariates, extra-demographic stochasticity, interpretation of parameter estimates.

6. Case study: Ebola. Model diagnostics and forecasting.

Additional background on time series analysis and POMP models, following the notation and approach of this short course, is provided in a full-semester extension of this couurse, Modeling and Analysis of Time Series Data.

This site is open source. Improve this page.