[From the syllabus for “POLI 229: Social Science Replication”, taught by Gareth Nellis at the University of California San Diego]
“The purpose of this class is to learn how to do cutting-edge empirical research in the social sciences by replicating others’ work. For each class, we will have a non-UCSD faculty member visit campus. A week before their visit, they will circulate a working paper—typically one drawing on an experimental or quasi-experimental design—along with the data and code needed to reproduce the results. Small teams of graduate students will reanalyze the data, and propose further robustness tests and extensions.”
“The class is aimed at students in the second, third, and fourth years of the PhD program in political science. A strong understanding of probability, regression, and causal inference is required. Prior knowledge of R is highly recommended.”
“The class tries to instill excellent workflow habits, and to teach tools that will make your research easier to conduct and more transparent. For this reason, the software and presentational requirements will be quite strictly enforced. All class work should be done using an R-RMarkdown-Git integration via RStudio and Github.”
“Replications, weekly memos, and pre-analysis plans should be written and coded in single, continuous RMarkdown (Rmd) documents. You should knit to html or pdf for written submissions, and to beamer or Xaringan for slide presentations. Everything should be “one-click” replicable, meaning all text and analysis output compiles from a raw Rmd with a single click of “Knit” in RStudio.”
“We will be getting to know DeclareDesign, a new R package that makes it straightforward to run Monte Carlo simulations of complex research designs, and thus to assess their properties (e.g. power, bias, RMSE). Follow this guide to get started, and play with the vignettes in the design library.”
To read the full syllabus, click here.