Analysis of the relationship between deaths from firearms, rates of gun ownership, and the ratio of mental health providers to the local population. This repository contains R scripts for the statistical analyses and the generation of pertinent tables and figures. Code is public, for manuscript review purposes, as of 18 May 2024.
.../R - Scripts
.../data - raw and cleaned data
.../results - tables, charts, figures, imputed data, etc
Script 1 - import and clean data
- Bins states into quartiles by gun ownership rate
- Generate contents of eTable 1
- Saves data as df.Rds
Script 2 - describe data
- Examine missingness
- Generate Table 1 and eTable 2
- Compare censored vs. uncensored subsets
Script 3 - imputation and modeling
- Pearson's correlations
- Figures 1 and 2
- Perform crude (unadjusted) and intermediate (limited adjustments) models
- Impute values for missing segregation data
- Apply a mixed effects model to the imputed data sets, and pool the coefficients into a final, complete-adjusted model.
- Generate Table 2 (concise and long form versions)
Script 4 - sensitivity analyses
- Compare random forest to primary mean matching, for imputation of segregation missing values
- Generate, visualize, and apply multiple sensitivity models for missing firearms numbers in censored counties
To quickly view the model results:
load("results/imputation.RData")
summary(mc4)
summary(mi.continuous)
summary(mi.qrt)
summary(pool_pmm) # full model, vs continuous values for firearm fatalities
summary(pool_pmm_qrt) # vs quartiled bins of firearm fatalities
To generate sensitivity model coefficients only, run the following parts of script 4
- Part 1
- Part 5, lines to create
sensitivity.list - Part 6 (helper functions)
- Part 7 (batch run of models)
- Kelly Corbett
- Malcolm Schongalla
- Erika Moen