Add warning when coordinates dataset contains both positive and negative z_stats

[JulioAPeraza]

Closes #465.

Changes proposed in this pull request:

• Check for z_stat in Dataset.coordinates.columns.
• Check if there are values with z_stat >= 0 AND z_stat < 0 in Dataset.__init__() and when the ImagesToCoordinates.transform() is applied.

[codecov]

Codecov Report

Merging #699 (3019aa9) into main (2b9229f) will increase coverage by 0.02%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main     #699      +/-   ##
==========================================
+ Coverage   85.31%   85.34%   +0.02%     
==========================================
  Files          41       41              
  Lines        4521     4530       +9     
==========================================
+ Hits         3857     3866       +9     
  Misses        664      664              

Impacted Files	Coverage Δ
nimare/dataset.py	89.96% <100.00%> (+0.23%)	⬆️
nimare/transforms.py	75.71% <100.00%> (+0.26%)	⬆️

---

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 2b9229f...3019aa9. Read the comment docs.

[tsalo]

It looks like the additions to transforms.py is covered by tests, but the changes to dataset.py are not.

[JulioAPeraza]

It looks like the additions to transforms.py is covered by tests, but the changes to dataset.py are not.

I think this happens because db_file = op.join(get_test_data_path(), "neurosynth_dset.json") contains Nones in the "z_stat" column, so it doesn't get to check for positive and negative values.

[tsalo]

Do you know how it's checked in transforms.py? Do we can any tests that actually cover the case where there are positive and negative z-stat values? My concern is that we don't currently have any methods that use z-stat values, so I think the only place we're going to catch any bugs is by testing the warning directly, rather than downstream in any meta-analyses.

[JulioAPeraza]

Do you know how it's checked in transforms.py?

I think it is checked when an images dataset is transformed to coordinates here: https://github.com/neurostuff/NiMARE/blob/main/nimare/tests/test_transforms.py#L131. We don't initialize a Dataset class here, so the changes to dataset.py are not covered.

I think to test the changes to dataset.py we would need to initialize the Dataset class with a source that contains a z_stat with negative and positive values.

[tsalo]

I think to test the changes to dataset.py we would need to initialize the Dataset class with a source that contains a z_stat with negative and positive values.

Maybe we can add z-stats to one of the test Datasets.

[JulioAPeraza]

Maybe we can add z-stats to one of the test Datasets.

I have added the new test, and it looks like it's covering dataset.py. After expanding the warning message I got an error in the Upload Coverage that I don't know how to fix.

[tsalo]

I re-ran that step and it worked this time. I assume it was just a random upload issue. I'll take another quick look and approve if it looks good.

[tsalo]

LGTM!

[JulioAPeraza]

Great. Thanks!

[tsalo]

@JulioAPeraza do you want to merge this?

[JulioAPeraza]

Sure!