This paper analyzes the likelihood of casualties in school shooting incidents across the United States using a Bayesian Logistic Regression model and historical data spanning from 1999 onward. By examining key predictors such as shooting type and geographic latitude, the study identifies factors that significantly influence casualty risks. The results provide actionable strategies for improving school safety, including enhanced threat assessment programs and stricter firearm regulations.
The persistent and devastating issue of school shootings in the United States necessitates a proactive, evidence-based approach. This project is designed to harness the power of predictive modeling to move beyond reactive policies, enabling more effective safety interventions and casualty prevention strategies for school communities.
data/raw_data: Contains the raw data obtained from The Washington Post. data/analysis_data: Contains the cleaned dataset prepared for analysis. model: Includes Bayesian Logistic Regression models and validation outputs. other: Houses supplementary materials, including notes and visualizations. paper: Contains the Quarto document, references, and final PDF of the paper. scripts: Contains R scripts for data downloading, cleaning, and modeling. Aspects of the code and the introduction were developed with the help of ChatGPT 4. Documentation of usage is available in inputs/llms/usage.txt.