jackonelli/covid_state_space_model

COVID-19 Modelling

Work in progress state space model of the corona pandemic.

Setup

Get the code

This repo contains a submodule post_lin_smooth, the easiest way to get the full code base is to recursively clone this repo by running:

git clone --recursive git@github.com:jackonelli/covid_state_space_model.git

Dependencies

Dependencies are listed in requirements.txt and can be installed with:

pip install -r requirements.txt

Get data from C19.se

Download data from https://c19.se/ to a local JSON file:

# Make sure that the current directory is the repo root
python src/c19_se.py --output-file <name-of_data-file>.json

Examples

The src directory contains example scripts to demonstrate the posterior linearization, they are supposed to be run from the repo root.

python src/<name-of-example>.py
# E.g.
python src/affine_example.py

TODO

• Setup CT test with tricky data, increased noise. Create situation where iterations are required.
• Iterative version of SLR smoothing.
• Naïve truncated sampling for FHM model(s)
• Gauss sampling on subspace (states sum to 1)
• Generalise SLR linearization to any(?) lin.
• Create a class for partial SLR.
• Toy SIR model, in place but trunc. sampling not working.
• Impl partial SLR for FHM model(s)
• Square root version of KF and RTS
• Actual FHM model
• Modified FHM model
• Parameter estimation (MCMC, possibly EM)