02.09.2014 Views

multivariate poisson hidden markov models for analysis of spatial ...

multivariate poisson hidden markov models for analysis of spatial ...

multivariate poisson hidden markov models for analysis of spatial ...

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.

focus <strong>of</strong> this section. Random fields, which possess a Markov property, are called<br />

Markov random fields (Elliott et al., 1995 and 1996).<br />

Let us generalize this idea in a two-dimensional setting. Let Z be the set <strong>of</strong> integers, and<br />

let<br />

2<br />

S ⊂ Z be a finite rectangular two-dimensional lattice <strong>of</strong> integer points. Typically, it<br />

will take S = { 0,1,..., n −1}<br />

× {0,1,..., m −1}<br />

, <strong>for</strong> some n and m . S is a two-dimensional<br />

lattice containing<br />

n × m points. The points in S are <strong>of</strong>ten called sites. To define a<br />

Markov structure on the set S , we define what is meant by two points being<br />

neighbours. Different definitions may suit different purposes, or applications. However,<br />

the following two general conditions should include in the definition.<br />

(i)<br />

A site must be a neighbour <strong>of</strong> itself.<br />

(ii) If t is neighbour <strong>of</strong> s , then s is a neighbour <strong>of</strong> t .<br />

The second condition is a symmetry requirement. It can be written<br />

s ~ t if the sites<br />

s , t<br />

∈<br />

S<br />

are neighbours. Two common neighbourhood structures are given in<br />

Figure 2.6. If s is a site, the neighbourhood<br />

Νs<br />

<strong>of</strong> s can be defined as the set <strong>of</strong> all its<br />

neighbours; Ν = { t∈S : t ~ s}.<br />

Hence, Figure 2.6 illustrates the neighbourhood <strong>of</strong> the<br />

s<br />

middle site, <strong>for</strong> two different structures. In these structures, special care must be taken at<br />

the edge <strong>of</strong> the lattice S , since sites located there have smaller neighbourhoods. One<br />

way <strong>of</strong> defining the neighbourhood structure is “wrapping around” the lattice and define<br />

sites at the other end <strong>of</strong> the lattice as neighbours.<br />

21

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

Saved successfully!

Ooh no, something went wrong!