add weighted quantiles

[gidden]

Description of PR

Adds a df.quantile() function which calculates (possibly weighted) quantiles for a variable across a population of models and scenarios.

If this feature is useful for others in the community, I'm happy to improve the PR, add a tutorial, etc.

cc @chrisroadmap

Please confirm that this PR has done the following:

• Tests Added
• Documentation Added
• [ ] Name of contributors Added to AUTHORS.rst
• Description in RELEASE_NOTES.md Added

TODOS - issues

• support custom names (currently put into model)
• more flexible weights argument

Adding to RELEASE_NOTES.md (remove section after adding to RELEASE_NOTES.md)

Please add a single line in the release notes similar to the following:

- (#XX)[http://link-to-pr.com] Added feature which does something

[codecov]

Codecov Report

Merging #686 (84c69ea) into main (947d8c4) will increase coverage by 0.0%.
The diff coverage is 93.2%.

@@          Coverage Diff          @@
##            main    #686   +/-   ##
=====================================
  Coverage   94.8%   94.9%           
=====================================
  Files         58      59    +1     
  Lines       5899    5990   +91     
=====================================
+ Hits        5598    5690   +92     
+ Misses       301     300    -1     

Impacted Files	Coverage Δ
pyam/iiasa.py	86.7% <ø> (ø)
tests/test_iiasa.py	97.8% <ø> (-0.1%)	⬇️
tests/test_io.py	98.4% <85.7%> (-1.6%)	⬇️
pyam/compute.py	96.2% <91.6%> (-3.8%)	⬇️
pyam/core.py	95.2% <100.0%> (+0.6%)	⬆️
pyam/utils.py	91.7% <100.0%> (ø)
tests/test_feature_quantiles.py	100.0% <100.0%> (ø)
tests/test_ops.py	100.0% <0.0%> (ø)
tests/test_core.py	100.0% <0.0%> (ø)
tests/test_feature_validation.py	100.0% <0.0%> (ø)
... and 2 more

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[danielhuppmann]

Thanks @gidden, this sounds really useful!

A few comments:

1. I developed a Statistics module for the SR15, see the docs and a real-world use case here. The main advantage of the Statistics class is that you often want to do quantiles on groups of scenarios (e.g., the temperature categorization). If I understand the current implementation correctly, the method will always compute quantiles grouped by scenario names indexed by model, region and variable - we should probably anticipate a more flexible approach.
2. If you proceed with implementation in this way, I suggest to add that as method like IamDataFrame.compute.quantiles() to avoid too many methods directly on the IamDataFrame class. See the docs.

[gidden]

@danielhuppmann - FYI - I am seeing test failures in test_datareader.py not related to this PR.

I'm also hoping to clean this up, time permitting, in the next days.

[danielhuppmann]

@danielhuppmann - FYI - I am seeing test failures in test_datareader.py not related to this PR.

But all checks except one passed, so I guess that the Woldbank-server just had a hick-up in that second. Re-running the failed test might do the trick...

[danielhuppmann]

Looks good, a few minor suggestions inline!

[danielhuppmann]

Three more minor fixes, then good to be merged. Thanks for adding a tutorial!

[gidden]

all committed, should be g2g @danielhuppmann ! Thanks for the review!

[danielhuppmann]

Release notes are still missing - pretty please!

[gidden]

Release notes updated and issues made @danielhuppmann

[danielhuppmann]

Thanks @gidden!