Skip to content
Prev 6959 / 20628 Next

mixed zero-inflated poisson regression

Hello,
You can easily program this in WinBUGS (in R via R2WinBUGS). You can 
then chose whether you want to model the binary part as a function of 
covariates or not. The problem is that you are increasing the complexity 
of the model...depending on your sample size, model complexity and 
strength of the relationships, you may encounter mixing trouble. The 
extra problem that you may have is that the number of fledgings tends to 
be small...and may be underdispersed...but this may depends on the 
species. In that case you have underdispersion and zero 
inflation....have fun.


Code for ZIP in WinBUGS can be found in for example Kery (2010), and can 
easily be extended for GLMMs. Alternatively, our next book is a 360 
monograph on zero inflated models for nested data (and/or with 
spatial/temporal correlation). See:   http://www.highstat.com/book4.htm  
Comes out late December, early January....just in case you need a 
Christmas present for yourself

Kind regards,

Alain