Kernel function and Kernel width

[Rodrigo Moncayo Estrada]

I received the following comments when applied NPMR with hyperniche. Please some help with the answer:

 

I was describing that: “Kernel functions are used to weight the observations around each predicted point, and the optimal kernel functions are selected by cross-validation”

 

A researcher asked: “The kernel function or the kernel width? Maybe both? Is only the kernel width optimized?”

 

Thank you in advance for your help

[Bruce McCune]

HyperNiche gives a choice of two kernel types (Gaussian or
rectangular) and two local models (local mean and local linear), but
the kernel width (smoothing parameter) itself is what is optimized. To
optimize the other choices one could try each of the four
combinations, but usually I find that I have decided in advance which
kernel type and local model makes the most sense for a given problem.
Bruce McCune

> --
> You received this message because you are subscribed to the Google Groups "HyperNiche and NPMR" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to hyperniche+...@googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.

[Rodrigo Moncayo Estrada]

We use the local mean Gaussian weighting function. Thank you for your answer Bruce