Skip to content
Prev 7297 / 20628 Next

Mixed model with zero truncated Poisson distributed data

[I'm taking the liberty of cc'ing this to the r-sig-mixed-models
mailing list, where it can be archived]
On 12-01-12 12:52 PM, Helen Ward wrote:
Here the problem is that you are specifying the random effect without
giving its name (i.e. random=~1|DadID), so R is trying to interpret it
as your desired link function (which is the fourth argument to the
glmmadmb() function: see ?glmmadmb).
As mentioned in another message on this list this morning, the problem
here is that you need to explicitly convert "DadID" to a factor from a
numeric value [e.g. data$DadID <- factor(data$DadID)] -- the current
error message is not very informative.
At a *quick* glance this looks reasonable.  Try plot(model2$Sol) and
plot(model2$VCV) to see if the trace and density plots look reasonable
(compare them with the results from example("MCMCglmm"). You should take
a quick look at the Overview and CourseNotes vignettes (e.g.
vignette("Overview",package="MCMCglmm")) (the latter in particular is a
bit overwhelming but well worth digging into if you're going to use
these methods)