Special Topics in Political Science (490-0-21)

Topic

Replication

Instructors

Alexander Coppock

Meeting Info

Scott Hall 212: Thurs 9:30AM - 12:20PM

Overview of class

Reanalysis of the data produced by research designs implemented by other scholars is a core research activity for empirical social scientists. Reanalysis can clarify (to the reanalyst and to others) the inferential basis of claims made in published papers, especially when the details of the analysis procedure are vaguely communicated by original study authors. Reanalysis can generate assessments of “robustness,” i.e., how estimates of the same estimand vary depending on the choice of estimator. Reanalysis can produce estimates of new estimands not considered by study authors; rather than faulting authors for analyses they did not conduct, reanalysts can simply carry them out. A particularly useful application of reanalysis is “meta-reanalysis” in which many reanalyses are aggregated to assess the generalizability of research claims.

This course is geared toward graduate students in the social sciences (especially but not exclusively political science) who will write empirical dissertations. Inevitably, the previous scholarship on any topic will have generated an empirical record; graduate students who aim to join the world’s experts on that topic ought to have reanalyzed that empirical record themselves for synthesis in their dissertations. Each week, we will reanalyze excellent randomized controlled trials from across the three main empirical subfields of political science (American politics, comparative politics, and international relations). We focus on RCTs because the identification of causal estimands flows from the randomization, allowing us to focus on analysis choices rather than on debates about endogeneity. The skills developed in the course often apply to nonexperimental designs as well, but we will practice them on the epistemologically secure ground of randomized experiments.

Registration Requirements

Graduate students only

Learning Objectives

Data cleaning, data standardization, analysis workflows, experimental analysis, robustness testing

Teaching Method

Graduate-student led seminar

Evaluation Method

Students will be the "lead" reanalyst one week; this presentation serves as their final project. The rest of the quarter, students will complete short reanalysis problem sets on that week’s paper and come to class very prepared to discuss the paper in full.

Enrollment Requirements

Enrollment Requirements: Reserved for Graduate Students.