May I define a node first stochastic and later as data?
The problem is: I would like to nest two Poisson models on count data.
The first model would allow to estimate incident cancer cases for a whole
country from mortality figures and incidence data known for some small
areas.
The second model should use the estimated incident cases for the whole
country as data and allow to estimate future number of cases based on
observed trends.
I think I cannot define the node for incident cases initially as schocastic and
later as data, but perhaps some one knows a trick. Using a nested model
would allow that the distribution of the extrapolation estimates take into
account the uncertainty of the data since it also has been estimated
previously.
Any help would be appreciated.
Victor
=====================================================
Victor Moreno [log in to unmask]
SERC Tel: 34-93 260 78 12
Institut Catala d'Oncologia Fax: 34-93 260 77 87
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|