[R-sig-eco] spatstat: help with calculating number of clusters, cluster size

[Chris Gast]

Soumya Prasad <prasadsoumya at ...> writes:

> 
> Hi,
> 
> I am new to spatial statistics and have a few basic queries. I am
> analysing the spatial pattern of trees on a large plot. I wanted to
> know if there is there some way of obtaining estimates of the number
> of clusters and cluster-radius (in an iterative fashion) after fitting
> a spatial point process to observed data (through 'ppm')?
> 
> Thanks in advance,
> Soumya
> 


Hi Soumya,

[Having trouble with the website--sorry if this ends up as a double (or triple)
post.]

I'm not very experienced myself, but I have recently been using some R code to
perform some of these analyses.  Here is what I have come across:

If you believe the underlying clustering model is a "Thomas model", you can use
the thomas.estK() function to estimate parameters for this model, which are
estimates of the number of "parent sites", average offspring radius, and average
number of offspring.

There is another function, kppm, that can do similar things, (can fit a Thomas
and Matern model), but the data must be "unmarked" prior to its use (use the
unmark() function).  I'll note that the data I'm using (the built-in "brambles"
dataset) gives different results sometimes with kppm() and thomas.estK() when
using the Thomas pattern (not really sure why).

I don't explicitly know how to compare these models to see which is "better",
but I think some careful googling might solve that problem.


Hope that helps get you started,

Chris