[R] Spatial Statistics on Grids

[Peter Biber]

Thanks again David,

I apologize for not getting back to you immediately.

I do not want to cost too much of your time, but if it's ok for you, I will
try to clarify what I'm looking for:

Yes, I'm not looking for density estimates. I need to characterize spatial
_patterns_ of objects distributed across an area (I'm aware that the use of
the term 'distribution' is somewhat ambiguous in this context, sorry). In my
case, I do not know the exact position of the objects, I just know, if this
type of object is located on a certain grid cell or not. Thus, I want to
analyse the spatial pattern of the grid cells containing the objects.

What I'm looking for is very closely related to what can be found in the
CRAN Task View 'Analysis of Spatial Data' under the topic 'Point pattern
analysis' or to the package 'MarkedPointProcess'. Actually, years ago, when
I wasn't aware of R, I implemented similar things in PASCAL and C++.

And yes, I absolutely agree with your line of thought, that my current
problem (having grid cells, not points) would be solved with point-based
methods, if the cell size would be small enough compared to the spatial
scale of interest. However that's not the case and the facts play a role,
that cell midpoints have a minimum distance which is, in addition, different
for edge-to-edge-neighbours and corner-to-corner-neighbours (e.g. in a grid
with quadratic cells).

Maybe, some of the packages mentioned above contain convenient methods for
that, too. However I didn't find any clue about that. That's why I posted my
question to r-help. 

Maybe I should post my question better to R-sig-Geo (as proposed by Don
McQueen) or R-sig-ecology.

What do you think?

Best regards
Peter







-----Ursprüngliche Nachricht-----
Von: David Winsemius [mailto:dwinsemius at comcast.net] 
Gesendet: Dienstag, 23. Juni 2009 20:33
An: Peter Biber
Cc: r-help at R-project.org
Betreff: Re: AW: [R] Spatial Statistics on Grids


On Jun 23, 2009, at 1:29 PM, Peter Biber wrote:

> Thanks David,
>
> Actually, I googled quite some time before posting my question. I'm 
> also aware, that there are many packages for spatial data analysis.
> However, I
> found no hint if any of them contains methods for characterizing (and 
> I don't mean just plotting) the distribution (e.g. pure random, 
> clustered,
> regular) of grid cells (not point data) with R.
>
> I someone out there has used R for similar tasks, I still would 
> appreciate his/her hints/comments.

I still cannot tell what sort of methods you have considered and rejected. I
would have thought that any process examined on a computer would be done on
a "regular grid". Points are an abstract concept, grids are the implemented
reality. I would have thought that notions of 2D density, which are
undoubtedly present in every introductory source regarding spatial analysis,
would be the obvious first pass at answering your question. Apparently you
are not looking for density estimates?

>
> -----Ursprüngliche Nachricht-----
> Von: David Winsemius [mailto:dwinsemius at comcast.net]
> Gesendet: Dienstag, 23. Juni 2009 16:47
> An: Peter Biber
> Cc: r-help at r-project.org
> Betreff: Re: [R] Spatial Statistics on Grids
>
>
> On Jun 23, 2009, at 10:21 AM, Peter Biber wrote:
>
>> Dear colleagues,
>>
>> I'm searching for methods to analyze spatial distributions of cells 
>> with certain properties in a regular grid. For instance, consider a 
>> grid, where a part of the cells have the property "0", and the others 
>> have the property "1". I'm looking for statistics I could use for 
>> characterizing the spatial distribution of the "1-cells".
>>
>> Does anyone know about R-packages, where such methods are 
>> implemented?
>>

David Winsemius, MD
Heritage Laboratories
West Hartford, CT