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

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

p values <strong>of</strong> the goodness <strong>of</strong> fit <strong>of</strong> the <strong>models</strong> with some two-fold interactions and<br />

without any interaction do not differ very much. There<strong>for</strong>e, the latent variables<br />

X = X , X , X , X , X , ) are decided to keep in the model (i.e. use all two-fold<br />

(<br />

1 2 3 12 13<br />

X<br />

23<br />

interaction terms). The vector <strong>of</strong> parameters is now θ = ( θ1, θ2, θ3, θ12, θ13, θ23)<br />

and thus<br />

the following restricted covariance model can be <strong>for</strong>mulated:<br />

Y<br />

Y<br />

Y<br />

1<br />

2<br />

3<br />

=<br />

=<br />

=<br />

X<br />

X<br />

X<br />

1<br />

2<br />

3<br />

+<br />

+<br />

+<br />

X<br />

X<br />

X<br />

12<br />

12<br />

13<br />

+<br />

+<br />

+<br />

X<br />

X<br />

X<br />

13<br />

23<br />

.<br />

23<br />

6.4 Data <strong>analysis</strong><br />

In this section, the computational results <strong>of</strong> the <strong>multivariate</strong> Poisson finite mixture<br />

model and the <strong>multivariate</strong> Poisson <strong>hidden</strong> Markov model with the restricted covariance<br />

structure is discussed and compared with the results <strong>of</strong> the local independence model<br />

and the common covariance structure. The computational results <strong>of</strong> the fully saturated<br />

<strong>multivariate</strong> Poisson finite mixture model and the fully saturated <strong>multivariate</strong> Poisson<br />

<strong>hidden</strong> Markov model will not be discussed since there is no available method to<br />

estimate the parameters <strong>of</strong> the fully saturated <strong>multivariate</strong> Poisson model in a reliable<br />

way. As mentioned in section 5.1, the computation <strong>of</strong> the fully saturated model involves<br />

a great number <strong>of</strong> summations and parameters to be estimated and this remains a<br />

difficulty <strong>for</strong> calculation <strong>of</strong> the probability function. As a result, a comparison with the<br />

fully saturated covariance model cannot be made.<br />

114

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

Saved successfully!

Ooh no, something went wrong!