Project 2) Replicating Claims from 2024 BPD Annual Report

For this assignment, we will ask you to work in pairs to conduct a replication analysis of various claims made by the Berkeley Police Department in their 2024 Annual Report.

Learning Objectives

We want you to use this project as an opportunity to develop and gain experience around the 2 major pillars of this class:

  1. Collaboration on Complex Projects: Gain experience coordinating with your teammate to complete tasks with dependencies on each other.

  2. Reproducibility: Gain experience describing your work in accessibile formats and making it reproducible.

2024 BPD Annual Report

Every year, the Berkeley Police Department (BPD) publishes an Annual Report to document their measurable progress, and to communicate their activities and achievements around five major topics:

  1. Staffing & Workload
  2. Crime & Investigations
  3. Road Safety & Collisions
  4. Stop Data
  5. Use of Force

You can find the 2024 Berkeley Police Department Annual Report in the following link:

https://berkeleyca.gov/sites/default/files/documents/BPD%202024%20Annual%20Report.pdf

Data Sets for Project 2

Some of the data collected by BPD is shared to the public through their so-called Transparency Hub: https://bpd-transparency-initiative-berkeleypd.hub.arcgis.com/. This is an online public data portal maintained by BPD to share near-real-time policing metrics, crime trends, and operational records.

For this assignment, we want you to focus on some of the claims reported in the sections “Road Safety & Collisions” and “Stop Data”. You can find the associated CSV data sets in the Data Portal tab of the Transparency Hub: https://bpd-transparency-initiative-berkeleypd.hub.arcgis.com/search

  • Collision Data (2011 to Current)
  • Berkeley PD Stop Data (2024 - Current)

Replication Targets

Here are the claims (e.g. numeric summaries, tables, graphics) from the 2024 BPD Annual Report that you will attempt to reproduce:

  1. Table of “Collision Totals” (page 18)
  2. Table of “Top collision locations” (page 19)
  3. Bar chart of “Primary Collision Factors, 2024” (page 19)
  4. Bar chart of “Most frequent vehicle violations, 2024” (page 21)
  5. Numeric summaries in the first two paragraphs of the “Stop Data Report” section (page 22)
  6. Bar chart of “Discretionary Stop Demographics during Inter-twilight Period, 2024” (page 24)

Note that parts a)-c) involve Collision Data, whereas parts d)-f) involve Stop Data.

Deliverables

For this assignment, you will have a single GitHub repository for your group. Here’s a suggested file structure, but feel free to adjust the structure if you consider it necessary.

├── .gitignore
├── README.md
├── Makefile
├── environment.yml
├── data/           # raw data files, and data dictionaries
├── code/           # R/Python scripts or notebooks
├── outputs/        # images, tables, numeric summaries
└── report/         # main report (PDF) with embedded outputs

Your repository should contain the following:

  • data/ folder with raw CSV files, and also with their corresponding data dictionaries (aka codebooks). BTW: You will have to create the codebooks.

    • Optional: you can also include derived “clean” versions of the data sets, but we’ll leave this decision up to your discretion.
    Note

    We are asking you to store the raw data files since each of them has a fairly small size of around 4.0MB. However, keep mind that this will not always be possible with larger files.

  • code/ folder with R/Python scripts and/or qmd/ipynb notebooks. The number of files, the kind of files, as well as their purpose is up to you. Make sure to follow consistent naming conventions, and good habits for organizing their content.

  • outputs/ folder with the main reproduced images, tables, and numeric summaries. These are to be used in your report.

  • A report/ folder with a corresponding report document (in PDF format, 10 pages max) containing your replicated outputs (e.g. numeric summaries, tables, graphics). Indicate which elements you were able to successfully reproduce, and if applicable, which elements you failed to reproduce or struggled to fully reproduce, and the reason(s) for such a failure.

  • A Makefile that includes at least:

    • .PHONY targets all and clean
    • all rule to produce report in PDF format, and other targets of your consideration
    • clean rule to remove PDF report, and other targets of your consideration

  • A README.md describing the aims of the project, the source of the 2024 BPD Annual Report, the sources of raw data files, important details/descriptions about the project’s structure. Also, include a section of personal reflections about the things you learn while working on this assignment.

  • Add a Conda environment.yml

  • An ai_documentation.txt file where your group will put the name of the AI tool, its model version [e.g. , ChatGPT-4, OpenAI], the prompts, and the obtained output.

Other Considerations

We will grade commit mechanics (how changes are packaged and documented) alongside workflow integrity (how git was used as a tool for reproducible research, rather than a single bulk upload at 11:59 PM).

  • Needless to say that you should avoid “dumping” the entire project in 1 or 2 massive commits at the deadline.

  • Each commit should represent a coherent, single step (e.g., "Add download data script" or "Fix plot axis labels in figure 2") rather than bundling unrelated files (data, scripts, and report) into one giant commit.

  • Commits should show steady development over the project window.

  • Concise, descriptive first lines (50–72 characters) stating what changed and why (e.g., Add unit test for missing values in survey data).

  • Do not use vague messages like update, fixed code, changes, or final submission.

  • Use of branches to collaborate.