[ENH][DOC] local join count and LOSH statistics

[jeffcsauer]

Purpose of PR

This PR addresses bullet point 1, 5, and 6 of #61. Specifically, this PR adds four new functions to estimate:

• Local spatial heteroskedasticity (function name: losh)
• Local join counts - univariate (function name: Local_Join_Count)
• Local join counts - bivariate (function name: Local_Join_Count_BV)
• Local join counts - multivariate (function name: Local_Join_Count_MV)

Each function is written in the form of a scikit-learn style estimator. PEP8 formatting has been applied to all of the functions, although a handful of lines are left long due to readability.

Documentation

Two notebooks are included in the PR. Each notebook briefly reviews the core math of the statistic and explains how a user might deploy the statistic in a geospatial workflow. The included examples match those found in R spdep::LOSH and spdep::LOSH.cs as well as the GeoDa Local Join Counts tutorial.

All functions include docstrings and doctests that use built-in PySAL example datasets.

Tests

I have added in four .py test files that execute successfully on my fork. These new tests follow the lead of the existing tests, specifically test_moran.py for the LOSH function and test_join_counts.py for the local join count functions.

Notes for PR consideration

• Local join count functions are currently using the default n_jobs=1 to avoid issues fixed in potential merges. Users may enter values other than 1.
• One doctest dataset, commpop, appears to have a dead link at the moment. Having trouble loading this as a libpysal.example. Once this dataset is working the doctest can be updated. Currently hosting it on a personal Github link.

[codecov-commenter]

Codecov Report

Merging #139 into master will increase coverage by 1.88%.
The diff coverage is 95.48%.

@@            Coverage Diff             @@
##           master     #139      +/-   ##
==========================================
+ Coverage   77.61%   79.49%   +1.88%     
==========================================
  Files          25       33       +8     
  Lines        3274     3584     +310     
==========================================
+ Hits         2541     2849     +308     
- Misses        733      735       +2     

Impacted Files	Coverage Δ
esda/losh.py	90.24% <90.24%> (ø)
esda/tests/test_ljc.py	91.30% <91.30%> (ø)
esda/tests/test_ljc_mv.py	92.00% <92.00%> (ø)
esda/tests/test_losh.py	92.00% <92.00%> (ø)
esda/tests/test_ljc_bv.py	92.30% <92.30%> (ø)
esda/local_join_count_bv.py	97.33% <97.33%> (ø)
esda/local_join_count.py	100.00% <100.00%> (ø)
esda/local_join_count_mv.py	100.00% <100.00%> (ø)
... and 7 more

---

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 e660983...2eb0d37. Read the comment docs.

[ljwolf]

👏 this is amazing @jeffcsauer! well done! extremely minor API changes that mesh with the sklearn developer guide, but jeez, these are awesome!

[ljwolf]

This has been superseded by #145