A paper is no longer the only research output, and for a lot of work in this program it is not the most reusable one. Releasing the dataset, the derivatives, and the analysis code makes the work citable in its own right — and makes it possible for someone to build on rather than re-derive it.
The mechanics are straightforward and the sequencing is not: you want a persistent identifier (a DOI) attached to a specific version, so that a citation resolves to exactly what was used, not to whatever the repository contains today.
Related¶
Share your work — the pathway step
Project management — covers repository choice
DataLad — versioned datasets make versioned DOIs meaningful