This project examines racial disparities in educational disciplinary actions, and whether these disparities are associated with regional estimates of implicit and explicit biases. ### Data sources All Project Implicit data can be found on the [OSF page](https://osf.io/y9hiq/) for that project. The CRDC data was obtained from [here](https://www2.ed.gov/about/offices/list/ocr/docs/crdc-2013-14.html) and [here](https://www2.ed.gov/about/offices/list/ocr/docs/crdc-2015-16.html) for years 2013/2014 and 2015/2016 respectively. All covariate data was downloaded from [American Fact Finder](https://factfinder.census.gov/faces/nav/jsf/pages/index.xhtml) or the [National Archive for Criminal Justice Data](https://www.icpsr.umich.edu/icpsrweb/content/NACJD/index.html). The shapefiles for the map in the paper can be downloaded from the US Census ([counties](https://www.census.gov/geo/maps-data/data/cbf/cbf_counties.html); [states](https://www.census.gov/geo/maps-data/data/cbf/cbf_state.html)) The datafiles contained in the data directory here are compositions of these sources. The primary, full data files, `full_model_data.csv` and `full_model_data_2016.csv` contains all outcomes and covariates. The `county_means.csv`/`county_means_2016.csv` and `county_means_sexuality.csv`/`county_means_2016.csv` contain the bias estimates for racial and sexuality biases, respectively. The `county_teacher_means.csv` file contains racial bias estimates computed from teachers in Project Implicit data (these were not computed for 2015/2016). The `covariates.csv`/`covariates_2016.csv` file contains the county-level covariates, and `covariates_imputed.csv`/`covariates_imputed_2016.csv` contain county-level covariates after performing imputation. The `map_data.csv`/`map_data2016.csv` file was created to simplify the map-making process. ### Models All models were estimated using a consensus monte-carlo procedure (see Scott et al., 2016 for details). The shards for each fitted model are saved in the models folder. Aggregated models, parameter estimates, and aggregated model posterior distributions are each available in the models folder. ### Code, Manuscript & other Documents All written documents were composed as R Markdown files. All computations and figures are self-contained within the documents, though occasionally the .tex files were edited to improve formatting or to remain consistent with publisher guidelines. All filepaths will need to be changed to suit the user's directory structure. ### Interactive Maps The interactive maps are maintained here for archival purposes. They can be downloaded and run in any browser. Web-based versions can be found on the lead author's [personal web page](http://www.travisriddle.com/riddle_sinclair_maps/). ### Shiny App Parameter estimates for all models fit during this research can be accessed via a shiny app. The code and data needed to run this app is maintained here, but they can also be accessed online at https://triddle.shinyapps.io/riddle_sinclair/ **Preregistration note:** After freezing our preanalysis plan, we discovered that the statistical model was misspecified. The updated version of the preanalysis plan has resolved the problem. **Preregistration Edit (1/1/2017):** We have been unable to fit models described in the preanalysis plan that are free from pathologies. We have accordingly updated our analysis plan with a slightly different strategy.