Seasonal model of Sharma et al.: COVID-19 Nonpharmaceutical Interventions Effectiveness

This repository contains one part of the code used in the paper Gavenčiak et al.: Seasonal variation in SARS-CoV-2 transmission in temperate climates: A Bayesian modelling study in 143 European regions, PLOS Comp. Bio., 2022. The 2021 preprint can be found here.

This repository contains the seasonal variant of the model of Sharma et al. (2021), Understanding the effectiveness of government interventions in Europe’s second wave of COVID-19 and has been forked from MrinankSharma/COVID19NPISecondWave; please see that repo for further details.

For the seasonal variant of the model of Brauner et al. Inferring the effectiveness of government interventions against COVID-19, see the repository gavento/covid_seasonal_Brauner.

Data

The main data file used in the model is data/modelSharma_dataSharma.csv which is identical to data/all_merged_data_2021-01-22.csv from Sharma et al. except for number formatting and carrying several more (unused) features.

The data files data/modelSharma_dataSharma_countryMobility_*.csv are enriched with Google community mobility reports. The column Mobility decrease is a mean of indicated mobility categories remapped to range from 0.0 (no mobility) to 1.0 (pre-pandemic mobility), as described in the paper.

Running the model

Instructions for recent linux distributions (E.g. Ubuntu 20.04+)

• Install poetry in case you don't already have it (follow instructions at https://python-poetry.org for non-default install or config).

curl -sSL https://raw.githubusercontent.com/python-poetry/poetry/master/get-poetry.py | python - --version 1.1.6
source $HOME/.poetry/env

• Install dependencies into a poetry virtualenv (once)

poetry install

• Run all or selected the inferences

Adjust the number of parralel runs: each paralllel run uses 4 CPU cores.

# Basic seasonality and sensitivity analysis
poetry run python scripts/sensitivity_dispatcher.py --max_parallel_runs 4 --model_config modelSharma_dataSharma \
  --num_samples 1250 --num_chains 4 --num_warmup 250 \
  --categories seasonality_basic_R_normal_Sharma seasonality_maxRday_normal_Sharma \
  seasonality_maxRday_fixed_Sharma seasonality_local_Sharma basic_R_normal_Sharma default_Sharma

# Mobility sensitivity analysis
poetry run python scripts/sensitivity_dispatcher.py --max_parallel_runs 4 --model_config modelSharma_dataSharma_countryMobility1 \
  --num_samples 1250 --num_chains 4 --num_warmup 250 \
  --categories seasonality_basic_R_normal_Sharma seasonality_maxRday_normal_Sharma

• Use notebooks in notebooks/final_results to create the plots.

Changelog

• Preprint v1 (tag preprint-v1)

  • Add seasonality model (also added to upstream)
  • Customized and extended plotters
  • Runners and configs for sensitivity analyses
  • Fixes and updates

• Preprint v2 (tag preprint-v2)

  • Added mobility sensitivity analysis, data and plotters
  • Rebased to updates in MrinankSharma/COVID19NPISecondWave commit 8884dc8.
    • Now includes real-world data from MrinankSharma/COVID19NPISecondWave
  • Update docs and configs for easier reproduction

• Plos Comp. Bio v1 (tag submitted-1)

  • Add sensitivity analyses for the seasonal forcing model and seasoanlity interactions
  • Add plotters for the new analyses and the final plots
  • Update major dependencies (numpyro, JAX, arviz), update code to match

Questions?

Please email Tomáš Gavenčiak (gavento at ucw dot cz) or Mrinank Sharma (mrinank at robots dot ac dot uk, only inquiries regarding their code) for questions regarding the codebase.