This repository contains the code required to:
- Download and clean public COVID-19 surveillance data.
- Estimate the time-varying reproduction number.
- Project future hospital admissions assuming no changes in transmissibility beyond reductions in susceptibility due to infections.
- Project other dependent metrics such as fatalities, Hospital bed usage, and mechanical ventilation bed occupancy as discrete scaled convolutions of hospital admissions and estimated parametric delay distributions.
- Produce a real-time report summarising these outputs.
See below for how to run a manual update using the provided docker container.
-
Optionally: set up the prebuilt docker container using the instructions here. Alternatively the estimates can be created outside of docker but dependencies will have to be installed manually.
-
Run the following in the terminal (giving GitHub credentials when requested). On a 16 core cluster a full update should take approximately 3 hours. Either interactively in the container or programmatically.
bash bin/update-forecasts.sh
- Save the estimates, for example pushing to a clone of the repo in GitHub. For an example of CRON commands required to automate this process see here.
A summary report is available in the docs/
directory. The latest version can be accessed here. Its source is in reports/nowcast-and-projections/report.Rmd
.
Individual data files and figures are available in the following locations:
- Formatted nowcasts (Rt, growth rate, etc) can be found in
format-rt/data/all
with plots available informat-rt/figures
. - Formatted projections can be found in
format-forecast/data/all
with plots available informat-forecast/figures/snapshot/<target>/<date>/<target>-using-<predictor>.png
. Other plots summarising fitting and forecasting are available in theformat-forecast/figures
. - Plots summarising convolution parameter values over time are available in the
format-forecast/figures/parameters
folder though these require careful interpretation.
See bin/update-forecasts.sh
for links to the various subscripts for different tasks. Stepping through these files may also be required for debugging.