DOC: derivatives datasets vs derivatives data

[CPernet]

@robertoostenveld and I have discussed a lack of clarity between derivatives datasets vs derivatives data. The information is there, but we need to think how to explain better.

Example found on OpenNeuro (@effigies )
dataset_description.json - DatasetType set to derivatives
derivatives
|- sub-01
|- sub-02
sub-01
sub-02

This is DatasetType raw but has derivative data. Several datasets are now flagged as derivatives, suggesting we have a communication issue.

Proposal: https://bids-specification.readthedocs.io/en/stable/common-principles.html#storage-of-derived-datasets headers raw dataset with derivative data, derivatives dataset, non compliant derivative data

[yarikoptic]

Is that to prime discussion on "derivative BIDS" at coming meeting in Copenhagen to keep it a "theme topic"? ;)

I think this is largely a historical artifact of encouraging containing derivative datasets under derivatives/ of raw BIDS. Having said that, there is 1 derivative I started to place under derivatives/ which I think makes total sense to include -- output of bids-validator, which is a derivative worth shipping within "raw" dataset. See e.g. ds005256

But as to me -- "raw" BIDS dataset already 99% of the time contains "derivative" data. The fact that they included some more derivatives on top of the "raw" ones makes it just heavier, and even potentially nesting "derivative" datasets, but overall it remains "raw BIDS dataset" . But indeed we might want to document it better. I am not quite certain what do you mean by "headers" @CPernet?

Also relates to

• [ENH] Add "study" DatasetType to organize a collection of source and derivative datasets #1972

[CPernet]

I meant 'titles' in https://bids-specification.readthedocs.io/en/stable/common-principles.html#storage-of-derived-datasets which also appear in the table of content.
What about showing:

• a raw dataset with sources and derivatives, DatasetType: Raw
• a derivatives dataset, DatasetType: Derivatives
• a raw with non compliant derivatives data, DatasetType: Raw

pretty much shows that already, but being more explicit about the content of the dataset_description.json
(i can do it, just want your opinion, then you can review it)