First off, I didn't know whether to post this here or on SO. I tossed a coin and ended here.
Second, I'm trying to get a grip on the kde2d function in the MASS library. I've read the manual on kde2d and bandwidth.nrd (which kde2d uses to compute its "vector of bandwidths for x and y directions").
In the manual page for bandwidth.nrd it says that that function is "A well-supported rule-of-thumb for choosing the bandwidth of a Gaussian kernel density estimator". Why would I want to use this function specific?
Also, I'm doing a project on cell data and spatial data (i.e. GPS coordinates of phones and the cell tower they've been on). How well does bandwidth.nrd work with specifically this kind of data and why? Are there alternatives?
Lastly, I've found that the MASS implementation of bandwidth.nrd multiplies its result by 4 "for compatibility reasons". I found that dividing the result of kde2d by 4 gives me a better-looking image (i.e. I can see roads and highway interchanges, not so much when I use the result as-is).
