"Heat map" advice/opinion
Thu, 10/27/2016 - 11:44 pm
I have a set of 2D coords and I'd like to plot the density of these points as a "heat map" (colour-coded map of probability at every location within a boundary). In the past, I have done one of two things:
1. 2D histogram - sample number of points in a grid and use that for representation, but this can look blocky. I have around 200 points so reducing bins can look weird with smaller bin sizes.
2. ImageInterpolate (voronoi) - for each point, calculate the number of neighbouring points within some arbitrary distance and use these values as z to make a surface with
ImageInterpolatewith voronoi option.
I'm wondering if a better approach would be to make a 2D gauss at each location and sum these together (perhaps with smoothing) to make the representation.
I'm not sure how to go about doing this other approach. I've tried to investigate the new
StatsKDEoption in IP7 without much success - does it work on 2D waves?
If anyone has:
- an opinion of what is the best way to make a heat map in these circumstances
- advice on how to do the third option.
I'd love to hear it. Otherwise, feel free to tell me to take this to stackexchange instead!