Chapter 1 Motivating & Rationale

1.1 A Case for Reproducibility

1.1.1 How far to go in the quest for reproducibility?

1.2 What is Snakemake & Why Should you use it?

1.3 Why R?

1.4 Working Example: Replicating Mankiw, Romer and Weil’s 1992 QJE

Throughout our tutorial we are going to use a running example to illustrate the concepts we discuss.

1.5 The way forward

For the purpose of this tutorial we will focus on replicating the following aspects of the MRW paper:3

  • Regression Tables 1 and 2: Estimating the Textbook- and Augmented Solow Model
  • Figure 1: Unconditional Versus Conditional Convergence

To replicate these we will need to proceed as follows:

  1. Perform some data management
    • Prepare the data before we run regressions
  2. Do some analysis. For example, run regressions for:
    1. Different subsets of data
    2. Alternative econometric specifications
  3. Turn the statistical output of the regressions into a tabular format that we can insert into a document
  4. Construct a set of graphs
  5. Integrate the tables and graphs into a paper and a set of slides (optional)

We hope that these 5 steps look familiar - as they were designed to represent a simplifed workflow for an applied economist or social science researcher.

Before proceeding to understanding how to use Snakemake and R to construct a reproducible workflow, the next chapter first takes a deeper dive into the a protypical way to set up a research project on our computer.

Exercise: Your own project’s steps

Think about a project you are working on or have worked on in the past (it may be a Bachelor or Master’s thesis or a recent / active research project). Does your project fit into the 5 steps we described above? If not, what would you modify or add to our 5 steps? (Do you think this would destroy the general principles we will encourage over the next chapters?)

  1. A complete replication using the concepts presented in this tutorial is available here↩︎