How to organize sample-wise group-level calls?

[tsalo]

For large datasets, it seems likely that folks will want to aggregate summary measures of smaller samples within the dataset with the "group" level option. For example, if you have fMRIPost-AROMA derivatives from ABCD (lots of subjects!) and you want to use the "group" option to summarize the postprocessing results, you might want to just look at a subset of the subjects, like data from the first session only.

The current plan would create a group-level set of results in the fMRIPost-AROMA derivatives dataset where the participant-level results live, but that's a problem if you want to run the group-level workflow on multiple subsets.

Should we organize nipost group-level arguments as:

• app preproc-dset postproc-dset
• app postproc-dset group-postproc-dset

The former seems more BIDS-y, but the latter allows users to create multiple group-level subsets without overwriting previous runs.

[effigies]

If the raw dataset is no longer needed, then passing the derivative as the input dataset seems perfectly in-keeping with BIDS.

[tsalo]

I can imagine cases where we might want to plot QC metrics from the post-processing subject-level run against demographics from the raw dataset's participants.tsv file, but we could just expect users to patch those into a post-processing participants.tsv file.

I just want to make sure that it won't conflict too badly with the BIDS execution BEP. I know that's not going to be merged into BIDS, but I'd still like to follow it to the extent feasible.