[R] spatstat - Fitting a Strauss model with trend determined by kernel density smoother
Rolf Turner
r.turner at auckland.ac.nz
Sun Jul 15 23:03:43 CEST 2007
On 14/07/2007, at 2:51 AM, Alejandro Veen wrote:
> Dear r-help,
>
> I would like to use the 'ppm' function of the 'spatstat' package to
> fit a Strauss inhibition model. I understand that I can specify a
> parametric model for the "background" trend, but how would I specify a
> trend which is estimated using a Kernel density smoother?
>
> In particular, I would like to use the 'kde' function of the 'ks'
> package to estimate the "background" intensity and then use this as
> the trend for a Strauss inhibition process.
Questions about a specific contributed package should usually be
directed to the maintainers of
that package rather than to r-help.
To attempt to answer your question:
You need to convert your estimate of the background trend to an
***image***; see the
function im() in the spatstat package.
Or instead of using kde, you could use the ppp method for density()
which is provided in
spatstat; this methop returns an image. See the help for density.ppp
().
Now suppose that your point pattern is ``X'' and your estimate of
the trend is ``bgim''.
You can then fit the model you want via
fit <- ppm(X,~bgim,inter=Strauss(42),covariates=list(bgim=bgim))
Note that you have to specify the interaction radius for the Strauss
model (I have specified this
radius to be 42 in the forgoing example); this radius is an
``irregular'' parameter --- i.e. it does
not appear in exponential family form --- and hence is not estimated
by ppm(), at least not
directly.
cheers,
Rolf Turner
######################################################################
Attention:\ This e-mail message is privileged and confidenti...{{dropped}}
More information about the R-help
mailing list