- #1
Jack Miller
- 4
- 4
Hello everybody!
Since some time I am trying the estimate the density of a set of numbers (in my case the numbers are distances to some object from a laser scanner).
As I read, that the kernel density estimation technique is a basic approach for that kind of problem. Different Kernels can be applied, e.g. a Gaussian kernel.
However the Variance/Bandwidth must be estimated for a correct project of the density from the given data(-->here the set of distances)
There are some measures like the "mean integrated squared error". In that case a set of test bandwidth are calculated and the one with the lowest MISE is a good choice for the bandwidth.
However there are other methods like the "plug in" method which results to an optimal bandwidth. Does anybody knows how to use that "plug in" method or knows some other method? I just stucked and that point of estimated the optimal bandwidth.
Any help or comment, I would be really glad!
Since some time I am trying the estimate the density of a set of numbers (in my case the numbers are distances to some object from a laser scanner).
As I read, that the kernel density estimation technique is a basic approach for that kind of problem. Different Kernels can be applied, e.g. a Gaussian kernel.
However the Variance/Bandwidth must be estimated for a correct project of the density from the given data(-->here the set of distances)
There are some measures like the "mean integrated squared error". In that case a set of test bandwidth are calculated and the one with the lowest MISE is a good choice for the bandwidth.
However there are other methods like the "plug in" method which results to an optimal bandwidth. Does anybody knows how to use that "plug in" method or knows some other method? I just stucked and that point of estimated the optimal bandwidth.
Any help or comment, I would be really glad!