Why now?
For the past four years, I’ve taught a social science statistics class, “Applied Quantitative Research Design.” Without going into all the details, I teach causal inference (econometrics) and prediction, including difference-in-differences, fixed effects, survey sampling and weighting, using R and revolving around exemplary papers in political science and economics published around 2010-2023, at the level of Angrist and Pischke’s Mastering Metrics.
This year, I thought I’d document some of my reflections on the class and notes on what I’ve tried, especially how AI and LLMs should change student learning. I was fortunate to have the Yale’s Center for Teaching and Learning assign Tina Zeng ‘28 as a Pedagogy Partner to my class, and I thought this blog would be a good way to document our thoughts as a teaching team (which includes other teaching fellows who may make an appearance later).
The course runs from September to December 2026. Expand below for the course content and the grading criteria:
Course topics
- 1.1 Causality; 1.2 dplyr
- 2.1 Units; 2.2 RCT paper discussion
- 3.1 Omitted variable bias; 3.2 OVB paper discussion
- 4.1 Noncompliance and instrumental variables; 4.2 IV paper discussion
- 5.1 Visualization; 5.2 Two-way fixed effects
- 6.1 Review; 6.2 TWFE paper discussion
- 7.1 Uncertainty and regression discontinuity; 7.2 RD paper discussion
- 8.1 Prediction; 8.2 Shrinkage, 9.1 ML paper discussion;
- 9.2 Standard errors, 10.1 Survey weighting; 10.2 Survey paper discussion
- 11.1 Synthetic control; 11.2 Meta-analysis
- 12.2 AI agents, big data, and advanced coding
Grade components
(Exact weights differ for undergraduate and graduate students.)
- Attendance and pre-class exercises: 5%
- Oral understanding (mostly a oral exam): 25%
- Problem sets: 10%
- Two in-class exams: 30%
- Final project: 30%
AI Disclosure: I wrote the post, AI set up the blog within my jekyll website,
and it also suggested a few titles for this first post, from which I chose “Why now?”