Analysing spatial point patterns in R - CSIRO
Analysing spatial point patterns in R - CSIRO
Analysing spatial point patterns in R - CSIRO
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
31.4 Randomisation tests 197<br />
31.4 Randomisation tests<br />
Simulation envelopes of summary functions can be used to test various null hypotheses for<br />
marked <strong>po<strong>in</strong>t</strong> <strong>patterns</strong>.<br />
31.4.1 Poisson null<br />
The null hypothesis of a homogeneous Poisson marked <strong>po<strong>in</strong>t</strong> process can be tested by direct<br />
simulation, us<strong>in</strong>g envelope as before. For example, us<strong>in</strong>g the cross-type K function as the test<br />
statistic,<br />
> data(amacr<strong>in</strong>e)<br />
> E plot(E, ma<strong>in</strong> = "test of marked Poisson model")<br />
test of marked Poisson model<br />
K on, off(r)<br />
0.00 0.05 0.10 0.15 0.20<br />
obs<br />
theo<br />
hi<br />
lo<br />
0.00 0.05 0.10 0.15 0.20 0.25<br />
r (one unit = 662 microns)<br />
Notice that the arguments i and j here do not match any of the formal arguments of<br />
envelope, so they are passed toKcross. This has the effect of call<strong>in</strong>gKcross(X, i="on", j="off")<br />
for each of the simulated <strong>po<strong>in</strong>t</strong> <strong>patterns</strong> X. Each simulated pattern is generated by the homogeneous<br />
Poisson <strong>po<strong>in</strong>t</strong> process with <strong>in</strong>tensities estimated from the dataset amacr<strong>in</strong>e.<br />
31.4.2 Independence of components<br />
It’s also possible to test other null hypotheses by a randomisation test. We discussed two popular<br />
null hypotheses:<br />
random labell<strong>in</strong>g: given the locations X, the marks are conditionally <strong>in</strong>dependent and<br />
identically distributed;<br />
<strong>in</strong>dependence of components: the sub-processes X m of <strong>po<strong>in</strong>t</strong>s of each mark m, are <strong>in</strong>dependent<br />
<strong>po<strong>in</strong>t</strong> processes.<br />
In a randomisation test of the <strong>in</strong>dependence-of-components hypothesis, the simulated <strong>patterns</strong><br />
X are generated from the dataset by splitt<strong>in</strong>g the data <strong>in</strong>to sub-<strong>patterns</strong> of <strong>po<strong>in</strong>t</strong>s of one<br />
type, and randomly shift<strong>in</strong>g these sub-<strong>patterns</strong>, <strong>in</strong>dependently of each other. The shift<strong>in</strong>g is<br />
performed by rshift:<br />
> E