Reproducibility, Transparency, and Data Ethics

The question¶

• How do you create transparency and credibility when you cannot share the data you use?

• What kind of data can you share?

• Strategies for data collection, data use, and data sharing

In an ideal world, you use public-use data, your code is all out there, and everything is out there. That is what journals want, and that is what research ethics suggests you do. Much of that is well established, and I will not rehash it here. Instead, this book is about how to square that transparency with the fact that some piece of almost every analysis requires confidentiality.

Maybe you are running a survey, and you owe privacy to your respondents: you promised it to them, and you signed off on that promise. Maybe you are using administrative data that is not yours. Maybe you purchased data, and even if you would like to make it public, the company that sold it to you (or made it accessible to you) does not want you to. What to do in these situations is much less well established than what to do with public-use data, so treat what follows as the start of a conversation rather than a set of rules.

Who this is for¶

• Researchers who collect their own data

• Researchers who use confidential data

• Researchers who thought they used public data

That last group is larger than you might think. Data that are called “public” are often not entirely shareable, and some of what you believe to be public data may turn into something you have to worry about once legal frameworks such as the GDPR come into play.

Roadmap¶

• When can you share data? -- how to determine whether you can share, from licenses and legal restrictions

• Ethics review, consent, and data sharing -- what IRBs weigh, what to put in consent forms, and a summary of Meyer (2018)

• Do you have data you cannot share? -- practices to find out what is actually in your data

• Making non-shareable data shareable -- ways of treating the data so they become more shareable

• Publishing data securely, with controls -- if you cannot make the data less revealing, ways to control who accesses them

• See also -- including how to keep confidential data out of your code

• Bibliography

A recurring theme: all of this needs to be built in from the start. Whether you can share the data affects how you collect them, how you write your code, and even how much money you ask your funder for.

Companion materials¶

• Slides: presentation (PDF)

• Part of the Reproducibility curriculum