What is expected exactly in terms of reproducibility?

Computo is not just about publishing a notebook and proving that it can be compiled with CI! This part of the process is what we call “Editorial Reproducibility”. “Scientific” or “numerical” reproducibility of the analyses is also mandatory, on top of classical peer-review evaluation.

We don’t ask people reproducing their data… yet! We also don’t ask for “bit-wise computational” reproducibility (i.e. obtaining exactly the same results bit-by-bit) but rather a “statistical” reproducibility, i.e. obtaining results leading to the same conclusion, with potential non-significant statistical variability.