[R-sig-Geo] Cell Declustering

[Edzer J. Pebesma]

Stefano, sorry for my rudeness to forward your private reply to r-sig-geo.

A quick and dirty answer would be to work with grids, and e.g. retain a 
single value in each grid cell.  Does anyone more familiar with the 
point pattern theory know of a more elegant way that does not need all 
kind of ad hoc grid choices? I believe the problem is thinning of a data 
set in areas where data are redundant.
--
Edzer

Stefano Pegoretti wrote:
> Dear Edzer,
>     many thanks for your very quick answer! Here is my "problem": I 
> have a large dataset of georeferenced samples, and I want to get a 
> smaller subset that must reproduce the same spatial pattern of the 
> biggest one; say it in another way, I want to "through away" randomly 
> some data, but honouring the spatial distribution of the starting 
> dataset. I don't know it's clear enough... :-P
>
>     Have a nice evening!
>
> stefano
>
>
> Edzer J. Pebesma ha scritto:
>> Stefano,
>>
>> I thought cell declustering meant finding the size of the region of 
>> influence for an observation and using that as weight in further 
>> analysis. An approach would be using voronoi diagrams (package 
>> deldir, and read the list archives), another using the number of 
>> nearest cells based on a regular discretization of the study area; 
>> for the latter you could misuse package gstat, interpolate record 
>> number with nmax=1 and compute a table of the resulting "prediction" 
>> grid. Ask me if you need an example.
>>
>> The description you give sounds like spatial stratfied sampling, and 
>> can be accomplished by method spsample in package sp, type = 
>> "stratified".
>> -- 
>> Edzer
>>
>> Stefano Pegoretti wrote:
>>> Dear List,
>>>     I have to perform a "cell spatial declustering" on my radon data,
>>> i.e. divide the study domain into a defined number of cell and for each
>>> of them randomly extracts a defined number of samples: does anybody 
>>> know
>>> if there are function or packages in R to quickly to this?
>>>     Thanks, and have a good day!
>>>
>>>   
>>> ------------------------------------------------------------------------ 
>>>
>>>
>>> _______________________________________________
>>> R-sig-Geo mailing list
>>> R-sig-Geo at stat.math.ethz.ch
>>> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>>>