I noticed this problem when writing tests for the new numpy-based performant Ripley statistics. When using a shapely polygon's convex hull, I was getting basically the correct intensity, but when I was using scipy, I was getting extreme overestimates of the intensity.
I narrowed it down to an issue with how scipy.spatial.ConvexHull.area works scipy/scipy#12290.
Depending on the fix there we need to either:
move to using hull.volume when getting the area from a scipy convex hull if upstream decides to adopt docstring clarifications
keep using area if upstream decides to adopt the proposed changes for area/volume of 2-d shapes.
I noticed this problem when writing tests for the new
numpy
-based performant Ripley statistics. When using a shapely polygon's convex hull, I was getting basically the correct intensity, but when I was usingscipy
, I was getting extreme overestimates of the intensity.I narrowed it down to an issue with how
scipy.spatial.ConvexHull.area
works scipy/scipy#12290.Depending on the fix there we need to either:
hull.volume
when getting the area from a scipy convex hull if upstream decides to adopt docstring clarificationsarea
if upstream decides to adopt the proposed changes for area/volume of 2-d shapes.