This repository accompanies the preprint "A Blue Start: A large-scale pairwise and higher-order social network dataset" by Alyssa Smith, Ilya Amburg, Sagar Kumar, Brooke Foucault Welles, and Nicholas W. Landry. It provides all codes for reproducing the analyses and figures in the manuscript.

• data contains data on the network statistics plotted in Figs. 1-6 in the paper.
• figures contains the Figs. 1-6 in the paper.
• starter-code contains example code for loading the datasets into igraph and graph-tool.

The following scripts are used to generate statistics from the network data:

• get_degree_sequences.py: This script is used to get the in- and out-degree sequences for the following network; it saves them as .csv or .csv.gz files.
• get_component_sizes.py: This script obtains the sizes of the weakly and strongly connected components for the following network and saves them as gzipped CSV files.
• get_starterpack_clustering: This script is used to get node cluster labels based on the Leiden algorithm and then compute the normalized entropies of each edge.
• get_starterpack_k_core.py: This script performs a k-core decomposition of the starter packs.
• get_starterpack_kendall_tau.py: This script compares nodal rankings from the starter pack network and compares to the following network ranking using the Kendall Tau measure.
• get_starterpack_pair_cooccurrence.py: This script returns the distribution of two-node co-occurrence frequencies in the starter pack network.

python get_pair_co-occurrence.py --input_filepath "deidentified_starterpack_hif.json.gz" --max_pack_size 4069 --num_workers 10

• get_starterpack_pair_s_line_counts.py: This script returns the number of nodes and edges in the s-line graph for $s=1,2,\dots,345$. It can be run as follows:

python s_line_count.py --input_filepath "deidentified_starterpack_hif.json.gz" --smin 1 --smax 345 --output data/s_count.csv

• get_starterpack_stats.py: This script returns basic statistics of the starter pack network as a JSON file.

• plot_network_stats.ipynb: This notebook plots Fig. 4 and also prints the basic network statistics in a readable way.
• plot_starterpack_stats.ipynb: This notebook plots Figs. 1-3, 5 and also prints the basic starter pack statistics in a readable way.

• starter-code/graph-tool_load.py loads the node dataset as a polars dataframe, the starter pack dataset in XGI, and the following network in graph-tool. This code uses about 310 GB of RAM and takes about 2.75 hours to run.
• starter-code/igraph_load.py loads the node dataset as a polars dataframe, the starter pack dataset in XGI , and the following network in igraph. This code uses about 460 GB of RAM and took about 5.5 hours to run.

Start by downloading the data from the Social Media Archive @ ICPSR (SOMAR).