15 Open Code

If research code is entirely secondary, it should be cited rather than shared directly (see Section 16). However, this is likely to be rare, and research code that is modified in any way from the original should be both cited and shared.

15.1 Tips

15.1.1 Make it reproducible

Include all the external files (e.g., data files) needed to run it
Use relative paths so that it can run on any computer
Set a seed if your code uses simulations or random number generation

15.1.2 Make it clear

Include a README that explains how to run the code
Assume that the audience has varying technical expertise and doesn't necessarily know the conventions of the language you're using
Indicate which parts of your code produce any figure, table, or value in your manuscript
- Beginner: include comments in your code like # produces Figure 3.1
- Intermediate: Use code the generate the text of the results section so you only have one thing to copy and paste into the manuscript
- Advanced: Use code to generate the entire manuscript (e.g., using the papaja package)

15.1.3 Indicate software versions

Beginner: Include a text file with the info, e.g., devtools::session_info() in R, or requirements.txt in python
Intermediate: Or use dependency management, like renv or packrat in R
Advanced: Or use containers like Binder, Docker or CodeOcean for full reproducibility

15.1.4 Confirm reproducibility

Beginner: Access your shared materials from a new computer and run the code
Intermediate: Or ask a colleague to try running your code on their machine
Intermediate: Or set up more formal code review (Vable et al., 2021) in your group
Advanced: Or use a service like CodeCheck

NB: If your code takes a very long time to run, such as when you have extremely large datasets or are running simulations, you can include smaller test datasets or run a smaller number of replications and include code at the top of the script to toggle real and test data or low and high numbers of reps.

14 Open Data

16 Citation