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 ...
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
For the case <strong>of</strong> two categorical variables, each with two levels ( 2× 2 table with present<br />
and absent <strong>of</strong> the species), to evaluate if an association exists between the variables the<br />
following model can be used:<br />
A B AB<br />
ln( F ) = µ + γ + γ + γ . (5.35)<br />
ij i j ij<br />
ln( F ij<br />
) is the log <strong>of</strong> the expected cell frequency <strong>of</strong> the cases <strong>for</strong> cell i,<br />
j in the<br />
contingency table.<br />
µ is the overall mean <strong>of</strong> the natural log <strong>of</strong> the expected frequencies<br />
γ represent the ‘effects’, which the variables have on the cell frequencies<br />
A and B are two categorical variables<br />
i and j refer to the categories within the variables<br />
There<strong>for</strong>e:<br />
A<br />
γ<br />
i<br />
= the main effect <strong>for</strong> variable A<br />
B<br />
γ<br />
j<br />
= the main effect <strong>for</strong> variable B<br />
AB<br />
γ<br />
ij<br />
= the interaction effect <strong>for</strong> variables A and B.<br />
The model presented by equation (5.35) is called the saturated model. It includes all<br />
possible one-way and two-way effects. Given that the saturated model has the same<br />
number <strong>of</strong> effects as there are cells in the contingency table, the expected cell<br />
frequencies will always exactly match the observed frequencies, with no degrees <strong>of</strong><br />
freedom remaining (Agresti, 2002). To find a more parsimonious model that will isolate<br />
the effects best explaining the data, a non-saturated model must be discovered. This<br />
model could be achieved by setting some <strong>of</strong> the effect parameters to zero. For instance,<br />
105