search

NSAPH-Projects / school_gun_incidents

Estimate causal and associative effects of firearm-dealer-to-school proximity on school gun incidents in U.S. census tracts
MIT License
0 stars 0 forks source link

readme

Project: Distance Between Schools and Gun Retailers and Odds of School Gun Incidents in the United States

Authors: Falco Bargagli Stoffi, Michelle Qin

Contributors: Michelle Audirac, Nishtha Sardana

Created: June 2022

Overview: Causal inference to estimate causal and associative effects of average firearm-dealer-to-school proximity on occurrence of a school gun incident within a school-containing census tract in the United States.

How to Run the Code in This Repository

  1. Request data_input_private.zip from the authors. Unzip in the data/input/private/ folder.

  2. Generate the dataset used for analysis and Table 1 by running:

    cd code
Rscript --vanilla make_datasets.R
Rscript --vanilla make_table1.R

  3. Install the CausalGPS R package (use version 0.4.0 or higher).

  4. Decide whether to run the GPS matching model that uses Generalized Estimating Equations (GEE) for the outcome model.

    • Set run_gee_model = F or run_gee_model = T in causal_analyses.R accordingly.
    • GEE models may take up to 96 GB to run (and up to 250 GB for the "trim 1/99" sensitivity analysis).

  5. Perform all main analyses by running:

    bash associational_analyses.sh
bash causal_analyses.sh

  6. Consolidate and visualize the results by running:

    Rscript --vanilla read_and_plot_covariate_balance.R
Rscript --vanilla read_all_model_results.R
Rscript --vanilla plot_main_results_as_odds_ratio.R

  7. Perform additional sensitivity analyses by running (i) commented lines in associational_analyses.sh and causal_analyses.sh and (ii) code in the supplementary/ folder.

    • When selecting which code to run in supplementary/, you will again need to decide whether to run GEE models (for GPS matching models and for the state-level spatial confounding robustness check, denoted gee_associational_model).
    • If you choose to run the batch scripts in this repo, first create a logs/ folder within supplementary/ to save the output.

Contact Us

Michelle Qin (mqin8 [at] jh.edu), Falco Bargagli Stoffi (fbargaglistoffi [at] hsph.harvard.edu)