MNT: Pybids 0.7 compatibility

[effigies]

This branch tracks restoration of functionality to FitLins, using derivatives from fMRIPrep 1.2.1 and models updated according to the draft BIDS model spec.

The ds114 model will be updated as needed in bids-standard/bids-examples#130.

Needed modifications to fMRIPrep derivatives are being tracked in nipreps/fmriprep#1384.

[effigies]

Currently stalled out at bids-standard/pybids#284.

[adelavega]

Summary of conversation with Tal:

1. At higher-levels, it looks like to compute contrasts, you feed them into the design matrix as the "intercept" and then computer a one sample t-test on the results. Need to test that this works with a two sample t-test (represented as 1/-1 in BIDS StatsModels)
2. We need to add support for flexible # of higher-levels. Currently only a "fixed" group level is supported. This is necessary for models with more than 1 run per subject (so that estimates are averages for each subject before passing on to group level)
3. We could potentially use the same interface for 1st and higher levels. Need to think about a refactor down the line for this.

[effigies]

@adelavega Given that we can now run simple second-level models, should we green-light a pybids 0.7 release?

[adelavega]

Ummm... I think the functionality is all there, but are the docs up to date? It's a pretty big API change so that's an important piece

[adelavega]

Also, maybe it's time to merge this PR as well, and we can start a new base PR for adding more than 2 levels (I've started working on it a bit)

[effigies]

Okay, let me give it a cleanup review.

[adelavega]

The main thing missing from previous functionality is some of the plotting-- design matrix correlation plot (no obvious "EVs"), and contrast matrix plotting. I think it would be reasonable to raise issues for restoring the function of those two (I don't personally need them so not a high priority for me)

[effigies]

There are also some weirdnesses with file naming. e.g.

derivatives/fitlins/sub-01/ses-retest/sub-01_ses-retest_task-fingerfootlips_bold_contrast-[array([[ 0. , 0. , 0. , 0. , 0. , 0. , 0. , 0. , 0. , 0. , 0. ,0. , 0. , 1. , -0.5, -0.5, 0. , 0. , 0. , 0. , 0. , 0. ,0. ]])]_ortho.png

[effigies]

From a quick read through, some notes for myself.

[adelavega]

oh, I thought I fixed that. the issue is its using .format(contrast) instead of .format(contrast['name'])

[effigies]

Oh, I think you did. I still had some old outputs in my derivatives directory.