Matlab ksdensity for bivariate data with more evaluation points

Question

I'm running Matlab code for kernel density, i.e., [f,xi] = ksdensity(x), where x is a two column bivariate data. The resulting output f is the density vector, while xi is the meshgrid of evaluation points that is 30x30 in dimension. See the documentation here: Link.

I'm trying to increase number of evaluation points that I receive from this code. There is an option mentioned in the documentation called 'NumPoints' that is only applicable for univariate data. Is there an option or ways that I can increase the meshgrid points of evaluation points of bivariate data to, say, 100x100?

Answer 1

You need to use the optional second input argument pts to specify the range and number of the output points in your grid. See this example in the documentation. Depending on your input data, you could specify something like this:

pts = [linspace(min(x(:,1)),max(x(:,1)),1000).' linspace(min(x(:,2)),max(x(:,2)),1000).'];

Answer 2

The NumPoints is npoints in the ksdensity(). e.g., [f,xi] = ksdensity(x, 'npoints', 1000) will return 1000 points of xi and f.