Output RMSE map and time series for decay model fit

[tsalo]

Closes #861. @madisoth, what kind of outputs do you think would be useful? I was thinking (1) a mean RMSE map and (2) maybe a mean RMSE time series, so users could flag volumes with bad model fits?

I could probably convert the RMSE map to a more meaningful unit, like percent?

Changes proposed in this pull request:

• Add a function to calculate RMSE 4D time series from monoexponential model fit.

[codecov[bot]]

Codecov Report

Attention: Patch coverage is 94.25287% with 5 lines in your changes are missing coverage. Please review.

Project coverage is 89.91%. Comparing base (af5e99a) to head (730354a).

Files	Patch %	Lines
tedana/decay.py	93.33%	1 Missing and 1 partial :warning:
tedana/io.py	84.61%	1 Missing and 1 partial :warning:
tedana/workflows/tedana.py	80.00%	0 Missing and 1 partial :warning:

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1044      +/-   ##
==========================================
+ Coverage   89.80%   89.91%   +0.11%     
==========================================
  Files          26       26              
  Lines        3540     3621      +81     
  Branches      621      629       +8     
==========================================
+ Hits         3179     3256      +77     
- Misses        212      214       +2     
- Partials      149      151       +2     

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

[tsalo]

@madisoth gave me the thumbs up on this offline, so I'm opening it up for review.

[tsalo]

Here's the mean RMSE image for the 3-echo dataset:

Here's the timeseries plot (with +/- 1 SD) for the same dataset:

[tsalo]

The results are now substantially different:

[handwerkerd]

The results are now substantially different:

I do expect the values to be non-trivially lower because you were including voxels with 1 good echo, which meant fit coefficients of 0 (i.e. the residual error would be the entire time series).

[tsalo]

Ah, okay. That's a relief then.

[tsalo]

That's fixed in my most recent commit.

[tsalo]

@handwerkerd I think everything should be good now. Would you mind taking another look?

[tsalo]

Thanks @handwerkerd!

Here's what the updated 3-echo timeseries plot looks like, for posterity.