mvksdensity- how to choose bandwidth and bandwidth comprises of how many points?

Though effect of bandwidth is seen ,its unclear how many data points are covered for eg:when we use bandwidth=1.is it circular or gaussian cone when using a trivariate data?

Answer 1

[Raunak Gupta]

Hi,

From the documentation of mvksdensity, the Bandwidth parameter can be selected using the Silverman’s rule of thumb mentioned here which takes number of data points, number of dimensions and standard deviation for a dimension into consideration while giving an effective bandwidth for a dimension. As for the number of points comprises in Bandwidth, it is essentially a smoothing parameter and applied on a dimension of data, so it covers all the data points.

For the last question, bandwidth will not set the kernel function which can be circular or gaussian cone. You may use Kernel Name-Value pair for specifying the same.

[ammu v]

but in mvksdensity, opimal bandwidth is given by the function only for one dimension.In other cases user specifies bandwith for other dimensions