Move the desc- entity to the end of the file name.

[arokem]

See discussion in bids-standard/bids-specification#1895

[arokem]

Also, @tsalo proposes that we change desc-mean to desc-stat, which would gel more nicely with the functional derivatives BEP and @tsalo's own GLM derivatives BEP. If folks think that's agreeable, I can implement that here as well.

[tsalo]

Sorry I miscommunicated my idea. I mean stat-mean instead of desc-mean, as in BEP 012.

[arokem]

Of course - thanks for setting me straight. I am actually not sure that it entirely makes sense for the use-case shown here, because we have desc-mean in contrast to desc-merged and while stat-mean makes sense, I am not sure that stat-merged does.

[PeerHerholz]

Was the idea to only have stat- or desc- or could both be used, ie stat- for mean and desc- for merged given that the underlying data is obtained through different operations, one being a statistic while the other isn't?

Sorry if I missed something obvious.

[Lestropie]

The "_desc-merged" data here is the comprehensive data. One could think of it as a concatenation in a way, but it's not really an explicit operation, but rather the absence of an operation. So one could make the argument for eg.:

"sub-01_model-bs_param-vf_dwimap.nii.gz": "",
"sub-01_model-bs_param-vf_dwimap.json": "",
"sub-01_model-bs_param-vf_stat-mean_dwimap.nii.gz": "",
"sub-01_model-bs_param-vf_stat-mean_dwimap.json": "",

The latter image is the result of computing the mean statistic along a particular image axis of the former.

Where this would become problematic is in the context of inheritance. All metadata in sub-01_model-bs_param-vf_dwimap.json would be inherited by sub-01_model-bs_param-vf_stat-mean_dwimap.nii.gz; if any of those metadata were to be inappropriate (eg. stating which image axis the bootstrap realisations are stored across), then that would be erroneous. However if the solution going forward is to not utilise inheritance and just duplicate all redundant metadata across sidecar files, then that wouldn't be a problem.