Final Project

My sole hope for this project is for it to further your goals for taking this course. Let’s collaborate to make it a good experience for you.





Project Options

Option 1: Data analysis

Collaboration: Groups of up to 3. Individual work is fine.

Perform a causal analysis on a dataset of your choice.

  • The methods you use for analysis will vary depending on your research questions and the structure of your data (e.g., different methods for quasi-experimental designs vs. general observational studies).
    • You’ll plan your methods in collaboration with the instructor.
  • A key part of all analyses will be sensitivity analyses to assess robustness of results to points of uncertainty in the analysis process.

Resources for finding data:



Option 2: Blog posts

Collaboration: Individual only.

Write two blog posts explaining causal inference ideas to a general audience.

  • Post 1: A Tour of Causal Inference
    • Which of our course ideas resonated most with you?
    • Lead the reader through these topics in an engaging and cohesive way.
  • Post 2: Pick any media item that has interested you. Write a reaction to it / an analysis of it from a causal inference perspective.



Option 3: Learn an advanced topic

Collaboration: Groups of up to 3. Individual work is fine.

Dig deeper into existing course topics or learn a new topic. Examples could include:

  • Methods for transportability (generalizability) of effects
  • Interference
  • Details of methods for time-varying treatments
  • Specialized considerations for particular study designs
  • Doubly-robust estimation



Option 4: Other

If none of these options piques your interest, I’m happy to discuss alternatives with you. Some ideas:

  • Design a Shiny app to illustrate causal concepts to students
  • Write (part of) an R package for making it easier to work with causal graphs
  • Critique of applied research
    • This is a good option for those who would like to do a data analysis but cannot find adequate data to pursue their question.
    • Find and read papers that study a question of interest to you. Critique these papers from a causal inference lens. This will involve constructing your own causal graph to guide a critique of the authors’ data collection and analysis methods.
    • Discuss what remains uncertain in this line of research, and propose an analysis plan for a new causal study to rectify the limitations of prior research.





Deliverables

There is a lot of flexibility in the form that your project takes. Examples include:

  • A video presentation
  • A podcast-style recording
  • A set of blog posts
  • A project webpage on a personal website

Work with the instructor to determine the most suitable deliverable for your project, depending on the option you pick.





Timeline

  • Milestone 1: Project Proposal Choose your final project option, topic, and group by Friday, March 10 at midnight CST.
    • Data analysis option: Find a dataset and formulate causal question(s) that you want to answer.
    • Blog post option: Outline ideas for one post
    • Advanced topic option: Settle on an advanced topic, and do some preliminary research to identify key ideas within this topic.


  • Milestone 2: Friday, March 31
    • Data analysis option: Understand your data context well and construct an initial causal graph.
    • Blog post option: Brainstorm ideas or create an outline for one post.
    • Advanced topic option: Make progress in learning about one sub-area for your topic.
    • Other: Make progress appropriate to the scope of the project.


  • Milestone 3: Sunday, April 9 (Check in in-person or via an email exchange)
    • Data analysis option: Finalize your causal graph and construct visualizations to inform an appropriate outcome regression and IPW analysis.
      • Finalizing your causal graph does not require, for example, doing causal discovery by hand. You should think through your graph holistically along the lines of the flow discussed in the Drawing Causal Diagrams chapter of The Effect (the reading from HW4). You should feel satisfied that you have represented all relevant variables and the relationships between them.
    • Blog post option: Draft Post 1 (A tour of causal inference)
    • Advanced topic option: Make progress in learning about one sub-area for your topic.
    • Other: Make progress appropriate to the scope of the project.


  • Milestone 4: Schedule a time to meet with me during class time this week via Calendly. (You can choose an in-person or Zoom option on Calendly.)
    • Data analysis option: Based on your visualizations, construct an outcome regression model, and conduct an IPW analysis. Refer to your workflow write-up from HW4 to guide the steps taken here. Organize information on causal effect estimates (overall, and possibly within subgroups, if that is of interest) and confidence intervals.
    • Blog post option: Make substantial progress on Post 1 (A tour of causal inference)
    • Advanced topic option: Make progress in learning about one sub-area for your topic.
    • Other: Make progress appropriate to the scope of the project.


  • Milestone 5: Schedule a time to meet with me this week via Calendly to share updates on your progress. (You can choose an in-person or Zoom option on Calendly.)
    • Data analysis option: Conduct a sensitivity analysis for unmeasured confounding for your project, and complete a draft of your deliverable.
    • Blog post option: Finalize Post 1 (A tour of causal inference) and make substantial progress on Post 2.
    • Advanced topic option: Make progress on the goals you set with the instructor in the previous week, and complete a draft of your deliverable.
    • Other: Make progress appropriate to the scope of the project, and complete a draft of your deliverable.