[R-sig-ME] Error in glmmADMB

[Ben Bolker]

Boris Berkhout <b.berkhout at ...> writes:

> 
> Dear list members,
> 
> I am trying to fit a glmmADMB model on my data, but even after trying
> several things and reading different posts on forums I can't seem to fix
> the problem I am encountering. My response variable is the proportion of
> individuals surviving at a certain timepoint. The model looks like this:
> 
> glmmSing8 =
> > glmmadmb(survF~temperature*snail*tank*plate+(1|plate:replicate)+(1|week),
> > data=SurvSing8, family="binomial", zeroInflation=TRUE)
> >
> 

 If you want to fit a binomial model, you need to specify the
_number_ surviving and the total (denominator), as in

glmmadmb(cbind(nSurv,nTotal) ~ ....)

in your case this might? be cbind(dead_cc,total) ?

otherwise I'm afraid the results won't make sense.  (glmer() has syntax
proportion~..., weights=nTotal , so that you can use the proportion as
the response variable, but I haven't implemented a similar
syntax for glmmADMB ...

  Are you sure you need zero-inflation in the binomial?  Did you try
it first with zeroInflation=FALSE?

  I'm also afraid that with snail crossed with everything else you'll
be overfitting your model: what is

ncol(model.matrix(~temperature*snail*tank*plate,data=SurvSing8))

?

In this case you are fitting the _interaction_ between snail and
the three-way interaction of temperature, tank, plate.  Was each
snail really measured in all combinations of temperature, tank,
and plate???
 

 [snip]
 
> 'snail' is a fixed effect, because I want to distinguish between snails and
> 'plate' and 'tank' are fixed effects, because they only have two and three
> levels respectively. A summary of my data looks like this:
> 
> > summary(SurvSing8)
> >      snail      temperature    plate   tank         week
> > replicate        time
> >  Y33    : 59   Min.   :15.00   1:287   1:189   13     : 97   7      : 77
> > Min.   :8
> >  Y64    : 59   1st Qu.:15.00   2:285   2:218   16     : 92   8      : 75
> > 1st Qu.:8
> >  Y40    : 55   Median :20.00           3:165   5      : 81   2      : 74
> > Median :8
> >  Y48    : 53   Mean   :17.53                   6      : 73   3      : 73
> > Mean   :8
> >  Y30    : 45   3rd Qu.:20.00                   7      : 64   4      : 71
> > 3rd Qu.:8
> >  Y51    : 45   Max.   :20.00                   8      : 64   1      : 69
> > Max.   :8
> >  (Other):256                                   (Other):101
> > (Other):133
> >      total         dead_cc          survF
> >  Min.   :10.0   Min.   : 0.00   Min.   :0.0000
> >  1st Qu.:13.0   1st Qu.: 6.75   1st Qu.:0.3820
> >  Median :20.0   Median :10.00   Median :0.5410
> >  Mean   :22.9   Mean   :11.83   Mean   :0.5346
> >  3rd Qu.:29.0   3rd Qu.:16.00   3rd Qu.:0.7037
> >  Max.   :78.0   Max.   :53.00   Max.   :1.0000
> >
> 
> Kind regards,
> Boris Berkhout
> 
> 	[[alternative HTML version deleted]]
> 
>