[R-sig-Geo] Memory size issue and setting bounding box when conducting Kest on multiple tree species

[Rolf Turner]

On 26/09/11 17:11, Laura S wrote:
> Dear all:
>
> (1) When I try to run Kest(data.ppp, correction=isotropic, nlarge=Inf...) I
> get the following error:
> Error: cannot allocate vector of size 463.4 Mb
> How do you normally overcome this issue? I have a SpatialPointsDataFrame
> that I converted into a ppp, so there are multiple tree species (points)
> being analyzed using Kest.

Well,  basically if a data set is too big, it's too big.  There are 
limitations.
Computers, even today, do not have infinite capacity.

One idea might be to try subdividing your point pattern into a number
of smaller patterns.  I.e. subdivide your window into a grid of rectangles,
clip the pattern to each of the rectangles, and analyze each resulting
pattern separately.  Ideally there should be a basic consistency between
the results of the ``sub-analyses''.  If such consistency is lacking, then
``interesting'' issues are raised about what the overall pattern is trying
to tell you.
> (2) How you recommend I set the bounding box, so that it is the same for all
> species? It is important that the spatial pattern of each tree species be
> analyzed in the same sampling window.

I don't really understand this question.  Your pattern should be equipped
with a window (and the window should consist of the region in which
points were *looked for*.  See Adrian Baddeley's workshop notes,
http://www.csiro.au/resources/Spatial-Point-Patterns-in-R.html,
section 2.4, page 15).

Generally speaking it is better to have all species in a single pattern
with the species indicated by the *marks* component of the "ppp"
object, rather than having each species in a separate pattern.

     cheers,

         Rolf Turner