[jmafoster1]

Summary

This PR adds an "easy interpret" option for causal data adequacy that puts the absolute kurtosis value through a sigmoid function and converts it to a "percentage". Closes #412. Credit to @SylviaWhittle for the original idea. Hope this isn't stepping on your toes, I just thought it'd be good to get it in the paper (which I plan to write this week).

What I ended up doing was taking the absolute kurtosis value and putting that through a sigmoid such that kurtoses close to zero correspond to values close to 100 and kurtoses far from zero correspond to values close to zero. A kurtosis of -0.8 corresponds to 50, since a uniform distribution (which represents just about the worst case that it's possible to get here) has a kurtosis of -1.2. Using -0.8 roughly corresponds to a distribution that's visually "about half way" between a uniform distribution and the normal distribution that gives a perfect value (the scale is not linear).

Main file changes

causal_testing/testing/data_adequacy.py - added a simplified_kurtosis function.

I also added the option to flip between raw and simplified values in the dashboard (with simplified being the default) and added explanations in there.

The other changes are tests and docs.

Checklist

Before you mark your PR as ready for review, please ensure you have completed the following.

• Added unit tests for new functionality
• Properly documented any new code
• Updated Sphinx documentation in docs/ and checked for updated configuration/runtime options
• Added a jupyter notebook tutorial in docs/source/tutorials for any new functionality
• Verified local docs build with sphinx-build -W source/ build

[github-actions]

🦙 MegaLinter status: ⚠️ WARNING

Descriptor	Linter	Files	Fixed	Errors	Elapsed time
⚠️ PYTHON	black	35		1	0.89s
✅ PYTHON	pylint	35		0	5.44s

See detailed report in MegaLinter reports

MegaLinter is graciously provided by

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 97.95%. Comparing base (7188c3d) to head (3c494de).
⚠️ Report is 6 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #424      +/-   ##
==========================================
+ Coverage   97.94%   97.95%   +0.01%     
==========================================
  Files          29       29              
  Lines        1898     1910      +12     
==========================================
+ Hits         1859     1871      +12     
  Misses         39       39              

Files with missing lines	Coverage Δ
causal_testing/testing/data_adequacy.py	100.00% <100.00%> (ø)
causal_testing/visualisation/geometry.py	100.00% <ø> (ø)
causal_testing/visualisation/testing_dashboard.py	100.00% <ø> (ø)
...sal_testing/visualisation/visualisation_plotter.py	100.00% <100.00%> (ø)

---

Continue to review full report in Codecov by Harness.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 8abdd0c...3c494de. Read the comment docs.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.