[R-sig-Geo] test for CSR in three dimensions

[Tobias Verbeke]

Adrian Baddeley wrote:
> Tobias Verbeke writes:
> 
>  >  Is there any functionality available to test
>  >  CSR for a three-dimensional point pattern ?
> 
> In 'spatstat' you can use the functions K3est, G3est, F3est to estimate 
> the three-dimensional versions of the K-function, the nearest neighbour 
> distance function, and the empty space function. Using simulation 
> envelopes, these functions can form the basis of a Monte Carlo test for 
> CSR. A realisation of CSR in three dimensions can be simulated using 
> 'rpoispp3'.
> 
> Currently (in spatstat 1.18-0) the envelope command cannot be used 
> directly to generate simulation envelopes. However this will be possible 
> in the next version 1.18-1.

Wonderful !

Many thanks,
Tobias

> In spatstat 1.18-1 you will be able to type (if X is a 3D point pattern)
>       plot(envelope(X, K3est, nsim=19))
> but in spatstat 1.18-0 you would have to use code similar to the 
> suggestion of Marcelino de la Cruz as follows below (thanks, Marcelino!)
> 
> Adrian Baddeley
> 
> =============================================================
> 
> n <- nrow(coords(X))
> volume <- volume.box3(as.box3(X))
> lambda <- n / volume
> 
> 
> # observed isotropic  corrected K
> kobs <- K3est(X, rmax=1,nrval=101)$iso
> 
> #99 simulated CSR patterns
> 
> Xsim<- vector("list", 99)
> for  (i in 1: 99){
>     Xsim[[i]] <- rpoispp3(lambda, domain = as.box3(X))
> }
> 
> #simulated K functions
>   ksim <- sapply(Xsim, function(x) K3est(x, rmax=1,nrval=101)$iso)
> 
> # pointwise Monte Carlo test
> plot(0:100, kobs, xlab="r", ylab="K3est(r)", type="l")
> matplot(0:100,t(apply(ksim,1,range)), lty=2,col=2, type="l", add=T)
> 
> _______________________________________________
> R-sig-Geo mailing list
> R-sig-Geo at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
> 
>