[R-sig-Geo] Boundaries for kernel density estimation

[Michael Sumner]

The soap-film smoothing of Simon Wood might be of interest:

http://www.maths.bath.ac.uk/~sw283/simon/papers/soap.pdf

I thought this was in the package mgcv, but I cannot find it. I'm sure
it is available as an R package somewhere.

Cheers, Mike.


On Fri, Sep 30, 2011 at 6:44 AM, Rolf Turner <r.turner at auckland.ac.nz> wrote:
> On 30/09/11 06:30, Jennifer Crees wrote:
>>
>> Hi,
>>
>>
>>
>> I am carrying out kernel estimation on some species point data. However,
>> the data are on a continental scale (many of the points hug the southern
>> coastline of Europe) so that the analysis is including marine areas when
>> the window is constructed around the point themselves. Is it possible to
>> specify a boundary so that only the sampled terrestrial area will be
>> included? Or perhaps it would be easier to look at local intensity?
>
> If your boundary were all coastal, you could get what you want by
> setting edge=FALSE in your call to density() --- I *think*!!!
>
> However if some parts of your window boundary are coastline and
> some parts are not, it looks tricky to me.  One idea might be to do
> one intensity estimate with edge=FALSE, another with edge=TRUE,
> and then somehow do a weighted average of the two with weights
> determined by distance from coastline.  Points near the coastline
> would get a high weight assigned to the edge=FALSE estimate and
> a low weight to the edge=TRUE estimate.  And vice-versa.
>
> Won't be too easy to implement, and it might be a load of dingos'
> kidneys anyway --- it's just an idea off the top of my head.
>
> I've cc-ed this message to Adrian Baddeley who may have a more
> sound suggestion.  Here's hoping.
>
>    cheers,
>
>        Rolf Turner
>
> _______________________________________________
> R-sig-Geo mailing list
> R-sig-Geo at r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>



-- 
Michael Sumner
Institute for Marine and Antarctic Studies, University of Tasmania
Hobart, Australia
e-mail: mdsumner at gmail.com