Hi, I want to estimate a model of the following form:
y = a1W1y + X1beta1 + e
e = a2W2e + X2beta2 + nu
with y a vector of proportions or counts and W a matrix of weights.
In other words, autocorrelation occurs for the response variable y and for the
disturbance term e. Ideally, at either level multiple W matrices could be
included and X1 and X2 could be (partly) different (or X2 could be absent). But
the most important thing is to be able to include a count/proportion response
variable in the autocorrelation model.
Does anyone know of a function in a package to estimate such a model? I have
found some functions that capture small parts of this, but especially the option
of a "generalized" model with a count reponse hasn't shown up yet.
But perhaps it exists under names I am not familiar with.
thanks a lot,
Hank Stevens (PhD. student)
http://toolbar.Care2.com Make your computer carbon-neutral (free).
http://www.Care2.com Green Living, Human Rights and more - 7 million members!
