Again, not much research in the last couple of weeks (and a first gap in writing…). This time, because of NeuroHackademy, which was very rewarding and interesting, but more about that elsewhere.

In terms of R&D, the main event of last week was the release of a version 1.0 of DIPY. It’s been about a decade in the making and the last couple of months have been a bit of a whirlwind of development, removing old code, and deprecating old APIs.

I still remain unsatisfied with the state of our stats module. In the short term, we have decided to warn users about potential changes to come down the line, but I still feel that I don’t understand enough about all the potential uses that we might encounter for statistics on values extracted along the length of the fibers (tractometry). This means that pinning down an API that is sufficiently general (for parallel existing, and all future uses) is really difficult. This is something that we will also have to think about as we evolve pyAFQ and AFQ-Insight. In these projects we have, for now, decided to focus on the case in which all the information from a bundle is reduced to a 1-dimensional vector. This assumption is more valid in some places (where bundles are relatively compact linear things) than in others (sheet-like bundles). It is clear that this does provide useful information, but the question is whether it also induces a bias in the detectability of effects. A thorny issue.