mggg / COI-Aggregation

COI Aggregation

This folder contains all of the necessary files to replicate the following graphs depicting the efficacy of the region-aware varaint of ReCom with respect to a custom score.

The score is computed as follows: For a given partitition $P={P_j}_{j=1}^{m}$, inclusion threshold $t$ and communities of interest $C_1, \dots, C_n$, we compute the preservation score according to the formula

$$f(P, t) = \sum_{i=1}^n \max\big[f_1(C_i, P, t),f_2(C_i, P, t)\big]$$

with

$$ f_1(C_i, P, t) = \max_{j}\left(\mathbf{1}_{Pop(C_i\cap P_j) \geq t,Pop(C_i)}\right) $$

and

$$ f_2(C_i, P, t) = \sum_{j}\frac{Pop(C_i\cap P_j),\mathbf{1}_{Pop(C_i\cap P_j) \geq t,Pop(P_j)}}{Pop(C_i)} $$

where $Pop(C_i)$ is the population of community of interest $C_i$.

Replication

First the user will need to install the necessary python packages provided in the python_packages.txt file. This as simple as running

pip install -r python_packages.txt

To replicate these graphs from scratch, you may run the provided python scripts in the order

1. run_chains.py
2. score_jsonl_files.py
3. make_graphs.py

Note that the first script will take several hours to run. For the convenience of the reader, the output of run_chains.py is provided in the jsonl_files folder in the XBEN format. Any interested reader my run the ben_decompress_files.py script to get the data in the correct format. For more information on using this script, see the GerryTools doumentation for BEN. Any user familiar with cargo and Rust may also run the bash file provided after installing the binary-ensemble package.

⚠️ Warning ⚠️ The jsonl files provided in thejsonl_files folder will consume 37GB of disk space when decompressed.

For the convenience of the reader, the outputs of the score_jsonl_files.py script are provided in the csv_files folder so that the user only needs to run make_graphs.py to replicate the graphs. This should also make it easier to analyze the data should the user wish to do so.

5x5 COI into 20 districts

10x10 COI into 20 districts

5x25 COI into 50 districts

2p5 COI into 20 districts