[R] Spatstat: Mark correlation function

[Rolf Turner]

On 24/11/12 06:36, AMFTom wrote:
> I normally use the following code to create a figure displaying the mark
> correlation function for the point pattern process "A":
>
> M<-markcorr(A)
> plot(M)
>
> I have now started to use the following code to perform 1000 Monte Carlo
> simulations of Complete Spatial Randomness (CSR).  It is a Monte Carlo test
> based on envelopes of the Mark correlation function obtained from simulated
> point patterns, normally used for Ripley's K:
>
> ME<-envelope(A, markcorr, nsim = 1000)
>
> And I produce the figure below.
>
> My question is: Is this a justified use of nsim if the envelope is based on
> simulations of CSR?

     Justified?  In terms of what?
> Or should I display the Mark correlation function
> without the envelopes?

     "Should" according to what criteria?  You can do whatever you
     like.  It's a free country.  I think.

     As someone's signature file says "What is the problem that you
     are trying to solve?"  What information do you want from the
     markcorr() function?  What information do you want from the
     envelope?

     Note that the markcorr() function is designed to assess whether
     there is spatial association of the *marks*.  The assumption is that
     the observed process is stationary --- spatially homogeneous ---
     not that it is constant intensity Poisson.  Thus CSR may not be an
     appropriate basis on which to form the envelope.  (Especially if
     your observed pattern is *not* Poisson!)

     You might be better advised to use rlabel():

     E <- envelope(X,markcorr,nsim=1000,simulate=expression(rlabel(X)))

     BTW why nsim=1000?  This strikes me as overkill.  I would probably
     take nsim=99.  (And if you really want the larger order of magnitude,
     it would be better to use nsim=999; nicer p-values! :-) )

         cheers,

             Rolf Turner