[Contributor] Seeking reviewer for Sheather-Jones bandwidth in scipy.stats.gaussian_kde

Contributor & Development Discussion
1

Hi everyone,

I’m Jared (qzyu999 on GitHub), and I’m looking to implement the Sheather-Jones bandwidth selection method for scipy.stats.gaussian_kde.

This has been discussed in Issue #19330, and there is strong consensus that it would be a valuable addition over the current Scott/Silverman rules (potentially using it as the default).

I have a background in this area, having previously implemented this in R, while providing a rigorous mathematical derivation of the solution (as seen here). I am looking to port a NumPy-based version that can be used within the SciPy community.

I’m looking for a maintainer/reviewer who can provide guidance on:

  1. Preferred implementation details.

  2. Integration specifics within the existing _kde.py module.

Is anyone available to shepherd this PR?

Edit: I’ve since created a PR: ENH: add Sheather-Jones bandwidth selection for gaussian_kde by qzyu999 · Pull Request #24733 · scipy/scipy · GitHub

Thanks!

Jared Yu

2

I’ve created a PR for the feature, porting the existing R implementation to Python as can be seen here: ENH: add Sheather-Jones bandwidth selection for gaussian_kde by qzyu999 · Pull Request #24733 · scipy/scipy · GitHub

Edit: The discussion was flagged for some reason, not sure why.

5

You were flagged for review because you were a new poster and had several links. I’ve restored your posts.

1 Like