[R] glmmadmb help

[Ben Bolker]

chchjames <james.mccarthy <at> windowslive.com> writes:

[snip]

> I am using the alpha version of glmmadmb, and it works for most of the time
> except for one of my models. The weird thing is that it has worked before, a
> couple of months ago, and for some reason it won't now and nothing has
> changed.
> 
> The code is:
> nbin5<-glmmadmb(stainp~beetle.ev+Caged*Section/SegmentT+
> (1|Site)+(1|Log.code),data=dat1,family="nbinom")
> 
> And it thinks for a while before giving me the error:
> Error in glmmadmb(stainp ~ beetle.ev + Caged * Section/SegmentT + (1 |  : 
>   The function maximizer failed
> In addition: Warning message:
> running command './glmmadmb -maxfn 500' had status 1 

Have you tried running it with verbose=TRUE ?

> I have narrowed it down, and the factor it has trouble with is "beetle.ev".
> This is a column of binary data which is presence or absence of beetles.
> There are only 0's and 1's in this column. The really weird thing is that I
> have another column for a different beetle ("beetle"), binary also, which
> works fine!

  {I am one of the authors of glmmADMB}

  It is pretty much impossible that the same code worked before on
the same data and that truly *nothing* has changed.  There must (?)
have been some minor change to the code or to the data,
even something apparently innocuous (and possibly a forehead-slapper
when you do figure out what it was).

  This is not a reproducible example, so I can't really diagnose
it.  Running with verbose=TRUE and posting the (voluminous) output
would help, but the best thing would be if you can post the data
somewhere, or a similar data set that displays the same problem.

  glmmADMB is a somewhat special-purpose package (being built
on AD Model Builder); you might want to consider sending follow-ups
to the AD Model Builder mailing list.

  Ben Bolker