[ENH] add images_to_coordinates

[jdkent]

Closes #423 .

Changes proposed in this pull request:

• use transformer class to make ImageToCoordinates

• test ImageToCoordinates

• add notebook demonstrating use of ImageToCoordinates

• add z_to_p function for cases when z-maps are not available, but you still specify z_threshold as a z-value

• test z_to_p

• fix bug in dict_to_coordinates

• How should I handle circular dependencies for get_masker?

  • have a round about somewhat hacky solution

• how to handle missing required stat maps?

  • I decided to add a merge_strategy parameter with three different options for users to decide whether to overwrite existing coordinates, and to skip generating coordinates for study contrasts without the required z or p maps.

• somewhat related to the above point, if I include peak statistics for the coordinates that are generated, how should the coordinates that already exist (and do not have peak statistics reported) be represented? As NaNs?

  • Representing missing peak coordinates as NaNs is my current solution.

• how should negative and positive statistics be handled?

  • there is two_sided parameter now available thanks to @tsalo

• how should I represent metadata

  • have a coordinate_source column with either original or nimare (or None)

[codecov]

Codecov Report

Merging #446 (c09647f) into master (b52b5af) will decrease coverage by 0.01%.
The diff coverage is 85.61%.

@@            Coverage Diff             @@
##           master     #446      +/-   ##
==========================================
- Coverage   83.25%   83.23%   -0.02%     
==========================================
  Files          40       40              
  Lines        4024     4110      +86     
==========================================
+ Hits         3350     3421      +71     
- Misses        674      689      +15     

Impacted Files	Coverage Δ
nimare/meta/cbma/ale.py	94.05% <ø> (ø)
nimare/meta/cbma/base.py	93.67% <ø> (ø)
nimare/meta/cbma/mkda.py	96.21% <ø> (ø)
nimare/transforms.py	71.04% <81.81%> (+0.53%)	⬆️
nimare/utils.py	94.20% <91.66%> (-0.63%)	⬇️
nimare/generate.py	100.00% <100.00%> (ø)
nimare/meta/kernel.py	97.50% <100.00%> (-0.02%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update b52b5af...c09647f. Read the comment docs.

[jdkent]

I think this is ready for review, @tyarkoni and @tsalo.
My current biggest question is how to work around the circular imports, I shuffled some functions around, but my solution may be non-optimal.
Once the docs test pass, I'll link to the example notebook I created demonstrating usage.
Here's a notebook from a previous build

[tyarkoni]

Looks good! Only a couple of minor comments, one of which is largely obsoleted by our conversation.

[tyarkoni]

Looks like the docstring needs updating to reflect that it's now in a class

[jdkent]

I think with one final look over, this should be ready to merge.

[tsalo]

Everything is looking really good. Most of my comments are about docstrings, although there is one about whether metadata should be logged. Also, there is a conflict that needs to be addressed. Finally, can you remove [WIP] from the PR title since it's ready for review?

[jdkent]

several rebases later, I think I've covered @tsalo review comments, ready for another check before merging (make sure my rebases did not undo anything)

[jdkent]

@tsalo I think this is ready for merge (I just don't want to do it without your approval)

[tsalo]

Sorry yes I went through and resolve all of my outstanding comments. I was just planning to check things over one last time today, but honestly I'm happy to merge now and we can address any issues as they come up.