Implement Recent Research?

[mbehrisch]

Hi,

great library! Works like a charm!! Just for a discussion: Do you think about implementing some recent research work/improvements over scagnostics?

Some pointers here are:

• [Improving the Robustness of Scagnostics] http://www.yunhaiwang.net/infoVis2019/Scagnostics/figfiles/Scag.pdf
• [Skeleton-based Scagnostics] https://webspace.science.uu.nl/~telea001/uploads/PAPERS/InfoVis17/paper2.pdf

[dvdjlaw]

Thanks for sharing! Here's my takeaway from a brief review of the papers:

Improving the Robustness of Scagnostics appears to introduce incremental improvements over the current algorithm:

• Adaptive binning
• Hierarchical clustering
• Adjusted method(s) for calculating "outlying" and "clumpy"

It sounds like adaptive binning would be a quick win. The other two would be slightly more complicated to implement but still within scope of pyscagnostics.

Skeleton-based Scagnostics takes a different approach of comparing different sets of scatterplots via a dissimilarity metric rather than single-summary measures like "outlying" etc. I'm not sure how this integrates with or complements the approach by Wilkinson et al which is used in this package.

Is this something you'd be interested in helping design/contribute?