Quantifying clusters of 3D points for varying properties

Question

I am trying to find a way to quantify how localized or distributed 3D points (x,y,z) are based on their properties. In the picture below, I have a distribution of points (x,y,z) that are varying colour from blue to red. The goal is to find a quantitative way to determine if blue dots are more localized in one dataset than in another. The ultimate goal however, involves 4D data (x,y,z,color) and quantifying that one localization is more blue than another.

From my research, I'm thinking kernel density estimations, nearest neighbours or some other cluster analysis. I would appreciate any ideas or suggestions on which option may be best.

Answer 1

Not sure I fully understand, but three things come to my mind:

1. Calculate the mean position (center) of all eg blue dots and then average the distance from each blue dot to the blue center.
2. Define a radius and count for each blue point the number of other blue points (neighbours) in that given radius. From all n take the mean.
3. Determine for each blue point the distance to all other blue points and get the mean of all means.

Each measure you can compare between datasets or colors