Overview

It is easy to agree that research should be transparent and reproducible. But, what does this even mean? Let’s use this overview from the World Bank Development Impact Blog:

Here, we will focus especially on the second concept, computational reproducibility. This is basically saying: If you provide your data and code to someone else, this person can run your code and will yield exactly the same results as you did. This should be the very minimum requirement, but in practice, this often fails due to missing material or code which does not run (Gertler et al. 2018, DIME results). In this chapter, we will discuss which steps you can take such that your code is computational reproducible.

But good scientific practice should go beyond this: Your code should also be clear to others, such that your decisions and analytical steps are verifiable and others can actually work with your code, e.g., doing sensitivity analyses. Until now, there are no general standards how exactly transparent, verifiable code should look like. This chapter will give you an overview on what to keep in mind when coding, but cannot give you finite rules.

If you are interested in reproducing and replicating the work of others, have a look at the BITSS Guide and the Replication Wiki (actually hosted by the University of Göttingen). Also keep an eye on the courses offered.

References
Further reading