Minority-group incubators and majority-group reservoirs for climate change adaptation

This repository contains code and supporting documentation for the agent-based model analyzed in our paper, "Minority-group incubators and majority-group reservoirs for climate change adaptation".

Quick start

To get started, clone this repository, e.g., execute the following in the terminal:

git clone https://github.com/eehh-stanford/SustainableCBA.git

After cloning the repository, install all dependencies by first starting the Julia REPL then run

julia> using Pkg; Pkg.activate("."); Pkg.instantiate()

Unit tests

We developed our model using test-driven development, which uses small, executable code snippets to confirm the model works as expected and to document model mechanics; see src/test/model.jl to view the test suite.

While still in the REPL, run the unit tests to make sure all is working well:

julia> include("src/test/model.jl")

This should print two "Test Summary" outputs where all tests are shown to pass. The tests initialize specially-initialized models and checks that model outputs are as expected.

Run the model and analyze results

Model and computational experiments

The model is implemented in src/model.jl and the computational experiments that run the model over all parameter settings for the desired number of trials and used by the Slurm scripts (below) is in src/experiment.jl.

Run all simulations on Slurm cluster

To run simulations on a Slurm cluster, log in to the cluster then execute the following commands from the project directory, first

./scripts/slurm/main.sh

to run the main analyses, and

./scripts/slurm/supplement.sh

to run the supplemental analyses. This creates a fresh, distinct version of simulation results that can be analyzed as we explain below, using archived data of the simulations used to create our results in the submitted version of the paper.

Analysis

With the addition of a network-theoretic modeling option there are two forms of analysis: (1) processing and calculating success rates and time to fixation from ABM output data, and (2) analyzing the frequency of network ties given different parameter settings.

Success rates and times to fixation

Use main_asymm_heatmaps to create the main heatmap results of success rate as a function of $h_\mathrm{min}$ and $h_\mathrm{maj}$, which can be found in scripts/plot.R. For creating the heatmaps of average time to model fixation, pass the keyword argument measure = "step" to main_asymm_heatmaps. Similarly, to create supplemental analyses use the supp_asymm_heatmaps function in scripts/plot.R.

To create the heatmaps you need the output data from the simulations presented in our journal article, stored in the data folder in the root project directory. To get the data in the right place, first create a data directory, then download and unzip the two zip files in our OSF repository: https://osf.io/cd9hx/.

To create time series of individual model runs, use the make_all_group_prevalence_comparisons function in scripts/analysis.jl.

Network tie frequencies

To understand the minority-group incubator and majority-group reservoir effects, we calculated the probability that a minority group member teaches at least one majority group member (currently Equation 6) and the probability that a majority group member teaches another majority group member (currently Equation 9). These probabilities are functions of minoirty group fraction, $m$, and majority group homophily, $h_{\mathrm{maj}}$. We inspect plots of each partial function individually. Plots over $m$ are made with the plot_over_m function, and plots over $h$ are made with the plot_over_h function. Both functions are contained in the file scripts/graph_counting.R.