How to create a SpatialMultiPoints object

On Mon, 14 Apr 2014, Stefan Jelkovich wrote:

Roger,

you are right. In case of the meuse data it makes no sense at all. Not
the best example... I do not see a clear adventage for the use of
MultiPoints, too. Maybe in case of categorical data? I only wanted to
create such an object to investigate the structure and prepare my
functions for MultiPoint objects.

So MultiPoints are not supported by the SpatialPoints and
SpatialPointsDataFrame classes? That's fine for me.

By the way: if MultiPoint objects are imported in R and converted to
SpatialPoints, the MultiPoint structure is just gone?

library(rgeos)
g4=readWKT("MULTIPOINT((3.5 5.6),(4.8 10.5))")
g4
class(g4)
str(g4)
slot(g4, "coords")

(they get identical rownames in the coordinate matrix, but by default
there is no key/ID checking for points, so this is not reliable). The
fact that the points belong to a multipoint geometry is not recorded
anywhere as it is an attribute of the points.

In your use case, the categorical observations have point support, not
multipoint support, so grouping points with the same category level is a
choice the user makes, but does confuse support. It could be done, but
would involve a lot of shoehorning to associate the points with their
appropriate geometry, and makes no great sense.

Roger

Thanks
Stefan

Am 13.04.2014 um 20:57 schrieb Roger Bivand <Roger.Bivand at nhh.no>:

On Sun, 13 Apr 2014, Stefan Jelkovich wrote:

Hi all,

I?m struggling with the creation of a
SpatialPoints/SpatialPointsDataFrame object containing MultiPoints.
Is this possible at all?

Why would it be possible, as each geomtry feature (point) is
associated with unique observed data values? A multipoint object is
of little use in point pattern analysis or geostatistics (or other
areas of spatial statistics) as its support is hard to use - the
single unique observed data value for each geometry would be
associated with multiple points, and a set of such objects might well
overlap mutually, meaning that it would be very hard to retrieve
observations.

If you feel that multipoint objects have a future in spatial
statistics, please contribute a suitable class container and methods
for handling it.

Hope this clarifies,

Roger

# SinglePoints are that easy:
library(sp)
data(meuse)
coordinates(meuse) <- ~x+y

# MultiPolygons are that easy:
p1 <- Polygon(cbind(x=c(0, 0, 10, 10, 0), y=c(0, 10, 10, 0, 0)))
p2 <- Polygon(cbind(x=c(13, 13, 17, 17, 13), y=c(13, 17, 17, 13, 13)))
lp <- list(p1, p2)
pls <- Polygons(lp, ID="1")
mp <- SpatialPolygons(list(pls))

How can I put all the meuse data points in one MultiPoint object?

Stefan

-- 
Roger Bivand
Department of Economics, Norwegian School of Economics,
Helleveien 30, N-5045 Bergen, Norway.
voice: +47 55 95 93 55; fax +47 55 95 91 00
e-mail: Roger.Bivand at nhh.no