vornoi bugs

I am in the process of matching North Carolina congressional districts (CD) to their corresponding voronoi polygons drawn from the CD centroids.

At the top, the voronoi polygons for the North Carolina CD centroids are calculated at an 80% buffer distance. This means that the algorithm is trying to extend the voronoi polygons out to 80% of the minimum bounding rectangle.

But, if you know things about voronoi polygons, you know that they

  1. always generate convex shapes
  2. extend an indifference frontier to infinity (i.e. the lines equidistant from the points being diagrammed continue forever if they do not intersect another indifference frontier)

But, for that southern part of the top map, we see that something’s screwed up.

Below, setting the buffer distance to 50% doesn’t have this happen.

There was apparently a related bug up on the QGIS project issue tracker, but this is unrelated to overlapping polygons, as it actually changes the angle of the indifference frontier between a few pairs of points. That bug may have also been present in ArcGIS current as of its posting, so maybe I can test against Arc 10.2.2 as well.

I bet PySAL’s got this on lock.

Now, I have to update QGIS, hope nothing breaks, and then try again. Maybe another reason why I wish my work computer was a rolling-release linux distro

imported from: yetanothergeographer