Adding PSI for continious data

[hakimelakhrass]

added Population Stability Index (PSI) for continuous data.

I used 0.25 as the alerting threshold citing this -> https://www.risk.net/journal-of-risk-model-validation/7725371/statistical-properties-of-the-population-stability-index#:~:text=In%20practice%2C%20the%20following%20%E2%80%9Crule,or%20type%20II%20error%20rates.

Used to Freedman-Diaconis Rule determine bin size

also updated the comments in drift/methods to say drift metric, instead of performance metrics

I recommend looking at the calculations and math closely to see whether it makes sense.

I'll add it for categorical next

[codecov]

Codecov Report

Attention: Patch coverage is 20.45455% with 35 lines in your changes are missing coverage. Please review.

Project coverage is 84.69%. Comparing base (dd20ef7) to head (c58c7f5).
Report is 75 commits behind head on main.

Files	Patch %	Lines
nannyml/drift/univariate/methods.py	20.45%	35 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #329      +/-   ##
==========================================
+ Coverage   83.40%   84.69%   +1.28%     
==========================================
  Files         100      100              
  Lines        7245     8931    +1686     
  Branches     1275     1730     +455     
==========================================
+ Hits         6043     7564    +1521     
- Misses        905     1016     +111     
- Partials      297      351      +54     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[nikml]

Small Comment:

You may want to consider numpy.histogram_bin_edges instead of manually implementing Freedman-Diaconis Rule. It has an option to use FD specifically as well, but maybe use doane like in JS. I remember we have looked into which option was the best back then.

We are also using it for JS here
https://github.com/NannyML/nannyml/blob/main/nannyml/drift/univariate/methods.py#L683

[nnansters]

Maybe also add some numerical tests in tests/drift/test_univariate_drift_methods.py?

This helps us verify that the behavior is correct now and doesn't change over time!

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.