Code repository: How contact patterns destabilize and modulate epidemic outbreaks

Article

• arXiv.2109.12180
• NJP

@misc{zierenberg_contact_patterns_2023,
	title = {How contact patterns destabilize and modulate epidemic outbreaks},
	url = {http://arxiv.org/abs/2109.12180},
	doi = {10.48550/arXiv.2109.12180},
	number = {{arXiv}:2109.12180},
	publisher = {{arXiv}},
	author = {Zierenberg, Johannes and Spitzner, F. Paul and Dehning, Jonas and Priesemann, Viola and Weigel, Martin and Wilczek, Michael},
	date = {2023-05-03},
	eprinttype = {arxiv},
	eprint = {2109.12180},
}

Data sources:

• Copenhagen Networks Study
• Sociopatterns

Prepare

Go to your directory

cd cloned_directory

Download the physical proximity data from the Copenhagen Networks Study

mkdir ./dat/
wget https://figshare.com/ndownloader/files/14000795 -O ./dat/bt_symmetric.csv

Create folders for output

mkdir ./out/

Running the analysis

Make sure you installed julia.

Start a julia REPL in the project folder

cd /path_to_cloned_directory

# Including run.jl will install required packages and provides easy acesse to functions to reproduce content of paper.
include("analysis/run.jl")

# To reproduce complete content for main paper run
reproduce_paper()

# For a more specific reproduction of individual content, follow the steps in reproduce_paper().
# We here provide an example for the data analysis (takes around ~6h)

# Set this to `true` to skip error estimates, as they take most of the time:
skip_jackknife = false

# You can reduce level of details to be faster but skip some analysis.
analyse_all(Copenhagen(), path_out = "./out/", level_of_details=3)

# You can filter out participants that had no rssi signal on both first and last day of study
analyse_all(Copenhagen(), path_out = "./out/", level_of_details=3,
    filter_out_incomplete=false)

# Repeat analysis for another dataset (e.g. InVS15)
analyse_all(InVS15(), path_out = "./out/", level_of_details=3)

Plotting

Plotting is implemented in python. It assumes that analysed files are placed in './out/' Install required packages, new conda enviornment recommended. Some smaller packages are only available via pip

conda install numpy scipy matplotlib seaborn h5py tqdm
pip install python-benedict addict palettable
pip install git+https://github.com/pSpitzner/bitsandbobs

Start an interactive python shell with our plot_helper

  cd resonance_contact_disease
  python -i ./plotting/plot_helper.py
  # or if you prefer ipython
  ipython -i ./plotting/plot_helper.py

We have some global settings that affect all panels.

# select things to draw for every panel
show_title = False
show_xlabel = False
show_ylabel = False
show_legend = False
show_legend_in_extra_panel = False
use_compact_size = True  # this recreates the small panel size of the manuscript
figures_only_to_disk = True
debug = False  # set to True to stop when a plot fails
figure_path = "./figs/njp"
data_input_path = "./out"

Note that, since we arranged panels in postprocessing we did not generate the labels and titles in matplotlib. Thus, there may be clipping of the automatically genereated axis labels and titles. They are still part of the pdf, just not in the viewer.

Load the main hdf5 file from the analysis. Then, figures can be created as shown below. They should open automatically, else try plt.show() to show them manually or plt.ion() to set matplotlib to interactive mode.

main_manuscript()
# or
figure_1()
figure_2()
figure_3()
figure_4()

To recreate SM figures see the figure_sm_ functions in /plotting/plot_helper.py.