On Fri, 12 May 2006, Xiaohua Dai wrote:
Please do not cross-post. Please also note that R has a list for spatial
data analysis questions, R-sig-geo. Try to decide first what your problem
is, then ask it on one relevant list only, trying later on another list if
no help is forthcoming. Why? Because the lists are archived, and the
archives are only a valuable resource if the record is near-complete. In
your case, fragments will be on several different lists.
I think your first question is "how do I find out which points belong to
an apriori unknown number of clusters in 2D?", and the answer might be
that this depends what you mean by a cluster for your phenomena of
interest.
Since we do not know much about your phenomena of interest (no informative
signature or affiliation), it isn't easy to see what you might think of as
appropriate criteria for deciding which points form a cluster, and which
might help in finding out how many clusters are present. Are you in
unexploded munitions, veterinary epidemiology, direct marketing, whatever?
The consequences of choosing appropriate cluster criteria are obvious, you
will know your own, but we do not.
When the points are assigned to clusters using your chosen criteria, in R
you might apply chull() in an *apply() call, but that is simple, and we
can continue this on R-sig-geo if you like. If a raster grid solution is
what you need - as you suggested initially, we can do that too.
You know your problem domain, you have to choose the criteria for what
constitute clusters first - and there are very many suitable functions in
R for doing that.
> Sorry, I did not make my question clear. Since I have a point theme
> with many points, some of them may clump together. the problems here
> are:
> 1. how to find clumps in a point theme?
> 2. the convex-hull extension I found only deal with all the points in
> a theme at each time? how to make each convex hull around each point
> clump automatically?
>
> Thanks.
>
> Xiaohua
>
>
>
> On 5/12/06, Bob Booth <bbooth at esri.com> wrote:
> > Xiaohua,
> >
> > That would be one way to do it. There are others.
> > Try searching ArcScripts for "convex hull"
> > http://arcscripts.esri.com/
> >
> > For example:
> > http://arcscripts.esri.com/details.asp?dbid=14535
> > or
> > http://arcscripts.esri.com/details.asp?dbid=12084
> >
> > Bob
> >
> >
> > -----Original Message-----
> > From: ESRI-L [mailto:ESRI-L at esri.com] On Behalf Of Xiaohua Dai
> > Sent: Friday, May 12, 2006 2:33 AM
> > To: ESRI-L at esri.com
> > Subject: [ESRI-L] outline polygons of point clumps
> >
> > Dear all,
> >
> > How to generate one outline polygon for each point clump? Are there
> > any present functions in ArcView, ArcGIS, R or some freewares? I just
> > had a quick look at the package adehabitat and did not find the
> > function.
> >
> > To my knowledge, I could do it as follows: 1) make a grid map of my
> > study area with cell values = 0; 2) assign 1 to the cells containing
> > at least one point; 3) convert the 1-value cells into polygons.
> >
> > Am I right?
> >
> > Thanks
> > Xiaohua
> >
--
Roger Bivand
Economic Geography Section, Department of Economics, Norwegian School of
Economics and Business Administration, Helleveien 30, N-5045 Bergen,
Norway. voice: +47 55 95 93 55; fax +47 55 95 95 43
e-mail: Roger.Bivand at nhh.no