29.01.2015 Views

Analysing spatial point patterns in R - CSIRO

Analysing spatial point patterns in R - CSIRO

Analysing spatial point patterns in R - CSIRO

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

26.4 Higher-order <strong>in</strong>teractions 159<br />

26.3.3 Other pairwise <strong>in</strong>teraction models<br />

Other pairwise <strong>in</strong>teractions that are considered <strong>in</strong> spatstat <strong>in</strong>clude the Strauss-hard core <strong>in</strong>teraction<br />

(with hard core distance h > 0 and <strong>in</strong>teraction distance r > h)<br />

⎧<br />

⎨ 0 if ||u − v|| ≤ h<br />

c(u,v) = γ if h < ||u − v|| ≤ r ,<br />

⎩<br />

1 if ||u − v|| > r<br />

the soft-core <strong>in</strong>teraction (with scale σ > 0 and <strong>in</strong>dex 0 < κ < 1)<br />

( )<br />

σ 2/κ<br />

c(u,v) =<br />

,<br />

||u − v||<br />

the Diggle-Gates-Stibbard <strong>in</strong>teraction (with <strong>in</strong>teraction range ρ)<br />

{ ( ) 2<br />

s<strong>in</strong> π||u−v||<br />

c(u,v) = 2ρ<br />

if ||u − v|| ≤ ρ<br />

1 if ||u − v|| > ρ ,<br />

the Diggle-Gratton <strong>in</strong>teraction (with hard core distance δ, <strong>in</strong>teraction distance ρ and <strong>in</strong>dex κ)<br />

⎧<br />

⎪⎨ (<br />

0<br />

)<br />

if ||u − v|| ≤ δ<br />

κ<br />

c(u,v) = ||u−v||−δ<br />

ρ−δ<br />

if δ < ||u − v|| ≤ ρ ,<br />

⎪⎩<br />

1 if ||u − v|| > ρ<br />

and the general piecewise constant <strong>in</strong>teraction <strong>in</strong> which c(||u − v||) is a step function of ||u − v||.<br />

<strong>in</strong>teraction<br />

0.0 0.2 0.4 0.6 0.8 1.0<br />

Piecewise constant <strong>in</strong>teraction<br />

0.00 0.05 0.10 0.15 0.20<br />

26.4 Higher-order <strong>in</strong>teractions<br />

There are some useful Gibbs <strong>po<strong>in</strong>t</strong> process models which exhibit <strong>in</strong>teractions of higher order,<br />

that is, <strong>in</strong> which the probability density has contributions from m-tuples of <strong>po<strong>in</strong>t</strong>s, where m > 2.<br />

One example is the area-<strong>in</strong>teraction or Widom-Rowl<strong>in</strong>son process [17] with probability density<br />

f(x) = αβ n(x) γ −A(x) (38)<br />

where α is the normalis<strong>in</strong>g constant, β > 0 is an <strong>in</strong>tensity parameter, and γ > 0 is an <strong>in</strong>teraction<br />

parameter. Here A(x) denotes the area of the region obta<strong>in</strong>ed by draw<strong>in</strong>g a disc of radius r<br />

centred at each <strong>po<strong>in</strong>t</strong> x i , and tak<strong>in</strong>g the union of these discs. The value γ = 1 aga<strong>in</strong> corresponds<br />

to a Poisson process, while γ < 1 produces a regular process and γ > 1 a clustered process.<br />

This process has <strong>in</strong>teractions of all orders. It can be used as a model for moderate regularity or<br />

cluster<strong>in</strong>g.<br />

Copyright<strong>CSIRO</strong> 2010

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!