ATM it is possible and legitimate (AFAIK, didn't check, bids-validator might have its own idea here on not having any sub-*/ , ref @effigies) to have a BIDS dataset
which would contain (raw) bids dataset and derivatives . They might even use datalad and YODA principles (so sourcedata/bids would also be a git submodule within derivatives/fmriprep-1 as it is e.g. in openneuro derivatives).
The question is either we want to define some dedicated DatasetType, e.g. "study", to make it explicit and then also possibly even "forbid" presence of individual sub- folders etc.
Somewhat also relates to BEP035 which defines compositions of datasets (asked the question in BEP035).
ATM it is possible and legitimate (AFAIK, didn't check, bids-validator might have its own idea here on not having any
sub-*/
, ref @effigies) to have a BIDS datasetwhich would contain (raw) bids dataset and derivatives . They might even use datalad and YODA principles (so
sourcedata/bids
would also be a git submodule withinderivatives/fmriprep-1
as it is e.g. in openneuro derivatives).The question is either we want to define some dedicated
DatasetType
, e.g."study"
, to make it explicit and then also possibly even "forbid" presence of individualsub-
folders etc.Somewhat also relates to BEP035 which defines compositions of datasets (asked the question in BEP035).
attn @jbpoline and @michellewang whom I have been grooming to adopt this mental paradigm in https://github.com/neurodatascience/nipoppy