scott cunningham

Course Description

This course provides a rigorous foundation in quantitative social science methods for first-year PhD students. After reviewing basic probability theory, we offer a systematic introduction to statistical inference and linear regression—the workhorse tools for empirical research in political science.

We take a "population-first" approach: define what you want to know about the population before worrying about estimation. Probability is the language for describing populations; statistics is the machinery for learning about them from samples.

Required Texts

• Blackwell, Matthew. A User's Guide to Statistical Inference and Regression. Free online
• Aronow, Peter M. and Benjamin T. Miller (A&M). Foundations of Agnostic Statistics. Cambridge University Press. Amazon

Suggested Text

• Angrist, Joshua D. and Jörn-Steffen Pischke (MHE). Mostly Harmless Econometrics: An Empiricist's Companion. Princeton University Press. Amazon — The most widely used text in applied causal inference. We reference Chapter 3 on regression. Inexpensive and worth owning.

Grading

Note: 70% of your grade comes from in-class exams. Problem sets are for learning; exams are for assessment.

Problem Sets

Problem sets are due Tuesdays at 11:59 PM. Each includes analytical problems and R simulation components. Weekly problem sets help you practice and internalize the material before exams.

Schedule

Part I: Statistical Inference (Weeks 1–7)

Part II: Inference & Regression (Weeks 8–13)

Practice Final Exam

The Practice Final Exam is a 15-problem packet organized into five scaffolded parts. Each part builds toward one of the five families of problem you will see on the actual final. The three problems within each part walk you up the skill stack: the first drills a single foundational move, the second combines two moves, and the third puts the whole stack together on a full in-family problem.

The point is not to memorize answers. The numbers, distributions, and design matrices on the real exam will be different. The point is to build pattern recognition: when you see an unfamiliar problem on the final, you should recognize which family it belongs to and know which tools to reach for. If you have practiced all three problems within a part, you should be able to execute the steps on a new problem in that family without hesitation.

Suggested approach. Do the third problem in each part (1c, 2c, 3a or 3b, 4c, 5c) first, closed-book and timed. The c-problems are the full rehearsals; the a's and b's are scaffolds you can fall back on if you get stuck. Solutions will be posted later this week.

Companion Shiny app for Problem 5c (variance-weighted regression): the OLS weights interactive app lets you slide $P(G{=}1)$, $p_0$, $p_1$, $\tau_0$, $\tau_1$ in real time and watch the ATE, the OLS coefficient, and the gap update live. The first load is slow (the WebAssembly R runtime downloads in your browser); after that it's instant. Use it to build intuition for how the variance weights respond to take-up and stratum size.

Course Policies

AI Policy

Do not use AI assistants (ChatGPT, Claude, Copilot, etc.) on problem sets. Work with your classmates instead. The learning happens when you struggle through confusion. The 70% of your grade that comes from in-class exams will reveal whether you actually understand the material.

Certain assignments in this course will permit or even encourage the use of generative artificial intelligence (GAI) tools such as ChatGPT. The default is that such use is disallowed unless otherwise stated. Any such use must be appropriately acknowledged and cited. It is each student’s responsibility to assess the validity and applicability of any GAI output that is submitted; you bear the final responsibility. Violations of this policy will be considered academic misconduct. We draw your attention to the fact that different classes at Harvard could implement different AI policies, and it is the student’s responsibility to conform to expectations for each course.

Collaboration

You may discuss problem sets with classmates, but you must write your own solutions and code independently. List all collaborators on your submission.

Late Policy

You have 2 late days total for the semester to use at your discretion—no approval needed. After your late days are used, problem sets lose 10% per day late. Additional extensions require advance approval.