graspp_2025_spring

Overview

• Run the code/notebooks in the cloud via Binder

  • 

• Course materials for "Data Science for Public Policy", a course at the University of Tokyo's Graduate School of Public Policy (Graspp)

• Instructor: Cory Baird

  • Github
  • Linkedin

Schedule

Module 1: How to Run Statistical Software (3 weeks)

• Week 1 (Apr. 7): The Easy Way to Code and Useful Tools
• Week 2 (Apr. 14): Acquiring Data through APIs
• Week 3 (Apr. 21): Downloading and transforming with tools (functions)

Module 2: Visualization (3 weeks)

• Week 4 (Apr. 28): Introduction to Data Visualization
• Week 5 (May 12): More visualization and mapping libraries
• Week 6 (May 19): Data pipeline and regression

Module 3: Regression, ML, AI

• Week 7 (May 26): Regression & Machine Learning
• Week 8 (June 2): ML & Neural Networks (A.I.)

Module 4: AI, LLM and Text analysis

• Week 9 (June 9): Scraping
• Week 10 (June 16): Reading PDF, NLP basics (Bag-of-words)
• Week 11 (June 23): Using LLMs
• Week 12 (June 30): Fine-tuning/training LLMs

Final Presentations

• Week 13 (July 7): Final presentations

Group Assignments/Milestones

• Milestone 1: Data selection and research question

  • Grade: 20% of grade
  • Task: Import and manipulate the data and show descriptive statistics in table or graphs.
  • Due: by Week 4 (Apr. 28)

• Milestone 2: Data Visulaization and Interpretation

  • Grade: 20% of grade
  • Task: Create at least 5 different visualizations (including charts) of the dataset.
  • Due: by Week 4 (May. 26)

• Milestone 3: Analytical Presentation

  • Grade: 20% of grade
  • Task: Present analysis in a whitepaper, slides or a dashboard
  • Due: by Week 11 (June 23)

Course Objectives

• Use Python to collect, clean, and analyze policy-relevant data.
• Design and implement reproducible research workflows to effectively manage and utilize public data.
• Apply statistical and machine learning methods to analyze policy problems
• Process and analyze text data using traditional NLP and modern LLMs (ChatGPT) to extract meaningful insights.
• Develop visualization to communicate research findings effectively to both technical and non-technical audiences.
• Collaborate effectively using professional data science tools like GitHub, Overleaf, and Google Colab.

Necessary software

• Code version control: Git/Github

  • GitHub Account: Create account then "star" class page
  • GitHub Desktop: For collaboration on code/notebooks
  • Git software: https://git-scm.com/downloads
    • git software is automatically downloaded with github desktop for mac but may not be for windows

• Running code AND notebooks

  • VSCode: For running notebooks and code (Download Link)
    • Sublime/PyCharm also acceptable
  • UV: Python version control and running notebooks (Download Link)

• If you are having issues running the previous software

  • The easiest way is to use github code space: This launches vscode in the cloud
  • Other solutions:
    • Anaconda: https://www.anaconda.com/
    • Jupyter.org Try: https://jupyter.org/try
    • Google Colab: https://colab.research.google.com/
    • pip: The standard tool for installing and managing extra Python libraries that provide specialized functions for data analysis, machine learning, and more.