This project is the result of the collaboration between the groups of Kai Nagel, TU Berlin, and Viola Priesemann, MPI for Dynamics and Self-Organisation, as part of the BMBF-funded infoXpand consortium. Main contributors: Emil Iftekhar, Sydyney Paltra, Jonas Dehning
This is the source code to analyse the various contributions that lead to changes in out-of-home duration in the year of 2020 compared to baseline years. The core method is Bayesian Inference using pymc.
To make the code run, it is necessary to download the covid19_inference module from github first: https://github.com/Priesemann-Group/covid19_inference/
The code is centered around main.py that uses functions in the following files:
- data_prep.py: Contains functions to prepare and return input data for the inference model
- model.py: Contains functions to build and return the inference model
- utils.py: Contains a few utility functions
- plot.py: Contains functions to plot the results the inference
- model_comparison.py: Contains functions to calculate the ELPDs
cluster_run.sh is used to run the inference on the cluster. Sometimes it is possible that no local data for the disease indicators is stored. As the cluster does not have internet access to download the Our World in Data data, one needs to run the cells of download_OWID.ipynb before using the cluster for the inference.
plot_results.ipynb is mainly used to plot model comparison figures. (It is quite messy right now.)
(input_data_analysis.ipynb is a messy notebook to pre-analyse the input data. May be archived at some point.)
- data: Storage for some of the (raw) input data
- results: Storage for the inference results, e.g. traces and source files
- figures: Storage for all the produced figures
- archived: old files