–Garret Christensen, BITSS Project Scientist
BITSS Bits. Too cheesy for our newsletter name? My guess is yes. Regardless, here’s an update on recent BITSS activities.
- I presented on p-hacking, publication bias, and pre-registration at the Making Social Science Transparent conference at the UC Davis Institute for Social Sciences April 22. My slides are here. There’s an excellent summary of the conference here, and videos of all the talks are here. My talk is embedded below.
And here’s Don Moore’s talk on how he’s working to change incentives through peer review and promotion and tenure review.
- I gave a workshop on reproducibility and reproducible workflows at the World Bank. A few of their data scientists have been itching for World Bank IT to approve use of Github, so we put together a daylong workshop with demos of how researchers there might benefit from Git/Github for version control and dynamic documents for automated document production and updating.
- David McKenzie of the World Bank Research Group and Jack Molyneaux of Millennium Challenge Corporation gave lightning talks on their work related to transparency and data sharing. David’s slides show his move Towards Reproducible Research over his career. No, he won’t be able to share the code and data from a paper he wrote 15+ years ago using Gauss, but pretty much all of his newer stuff is on his website and in the World Bank’s Microdata Catalog.
- We used the Github Desktop app to learn the first steps about version control. Github Desktop is a GUI app for version control. You can do version control from the command line, but I’ve started to prefer the GUI app for demos for simplicity’s sake. (Unless you’re on Linux, Github Desktop looks the same on Mac and Windows.) There are many GUI apps to choose from, and you might prefer another over Github Desktop. The unfortunate thing is that practically all of the online tutorials, including Software Carpentry and Atlassian are for the command line only.
- I demonstrated three ways to make dynamic documents. Files are here.
- Stata + LaTeX
- Stata + Markdoc
- R Markdown/R Studio
- R Studio is the only truly complete one-click reproducible workflow method of these, where I think you could write an entire paper and handle all your output, plus it’s got version control built in (Stata, let’s see this in Version 15, please!) and use can use the R Project functionality to manage directories and such. but the World Bank uses Stata and MS Word a lot, and there are ways to make that more reproducible. Markdoc is a user-written ado for Stata that can do a lot. (But can you format regression tables for MS Word using outreg or estout? Someone asked, and I don’t know a way to do that.)
- BITSS is hiring! We need a new program associate, as Alex Wais is leaving us for a cool research project on a United Nations project. Good luck Alex!
- Want to get in on some of the BITSS training similar to what I described above? We’ll be in Orlando in late June at a workshop before the American Library Association meeting, and we’re offering a free two-day training at ICPSR’s Summer Program at the University of Michigan in early July. Sign up now! We’ll also be at the Association for Psychological Science meeting in Chicago in late May if you want to stop by our booth and say hello.