RSH 10: Reproducibility

How to make research reproducible

Collaborative notes taken during the session

What tool/trick you recommend to make your work/research more reproducible?

Think of your future self, code+take notes so that your future self will understand what you did
Code with more than one person, joint projects
Some means to install software libraries needed
Conda environments!

Does reproducible also mean completely automated? If others have to be able to run the analysis, it feels like that.
- automation could be bad and not explicitly specify code versions that break an analysis
As a reviewer: do you ask for the code? Do you review the code?
- I reviewed once a data+software paper and told the editor that I can review the code if they give me 2 weeks more... never heard back :)
- I usually glance at the code. Once found a serious bug.
ReproHacks:
- https://twitter.com/reprohack
- https://reprohack.github.io/reprohack-hq/
How do tasks to make work reproducible change given the scale of the project? LHC analysis groups won't act like a small group of ecologists.
- an example: https://www.nature.com/articles/s41567-018-0342-2
ReproHack is a great idea: journal-club but for github repro's: try to run the so called "reproducible" code shared by the authors of a paper
- yes!
To make our research reproducible, do we have to be using notebooks (Rmarkdown, Jupyter, etc)?
- I personally think the other way around. Full automation (no interaction) -> more r