Which error family: beta or bionomial with weights?

[Jem_Powell]

Hello!

I'm investigating the effect of different weather variables (temperature, wind, and rain) on the breeding success of blue tits. My data for breeding success ("overall success") is a proportion bounded between 0 and 1, but also includes 0s and 1s. I was trying to use a glmmabmb with a beta error family, but with no success (model and error message below):

Overall.suc<- glmmadmb(Overall.success~ Av.Temp.Brood+Av.Wind.Brood+Av.Rain.Brood+
(1|NestBoxID),
family = "beta",link = "logit",
data = b_data)

*Error in glmmadmb(Overall.success ~ Av.Temp.Brood + Av.Wind.Brood + Av.Rain.Brood + : *

• The function maximizer failed (couldn't find parameter file) Troubleshooting steps include (1) run with 'save.dir' set and inspect output files; (2) change run parameters: see '?admbControl';(3) re-run with debug=TRUE for more information on failure mode*
  In addition: Warning message:
  In system(cmd, intern = intern, wait = wait | intern, show.output.on.console = wait, :
• running command 'C:\WINDOWS\system32\cmd.exe /c glmmadmb -maxfn 500 -maxph 5 -noinit -shess' had status 1*

I then tried switching instead to a glmer with binomial error family and weighting (nest clutch size in this case). This ran but I still got some error messages so I included a glmer control. This also ran but still with some warnings (shown below):

Overall.suc_GLMER<- glmer(Overall.success~ Av.Temp.Brood+Av.Wind.Brood+Av.Rain.Brood
+ (1|NestBoxID),
family= binomial(link= "probit"),
weights = Clutch.size, data = b_data,
glmerControl(optimizer = "bobyqa", optCtrl = list(maxfun = 100000)))
Warning message:
In eval(family$initialize, rho) : non-integer #successes in a binomial glm!

I was wondering if anyone had any alternative/ better suggestions of how to run this model? Ideally without any warning messages...

Thanks!
Jem

[nirgrahamuk]

if you ran it with family gaussian, is it significantly better or worse than the binomial ?

[Jem_Powell]

The data for overall success is not normally distributed at all so I didn't think gaussian would work (I've attached the histogram). When I run the model with gaussian I get the warning message shown below:

Overall.suc<- glmmadmb(Overall.success~ Av.Temp.Brood+Av.Wind.Brood+Av.Rain.Brood+

•                   (1|NestBoxID),
  

•                family = "gaussian",link = "identity",
  

•                data = b_data)
  

Warning message:
In eval(substitute(expr), data, enclos = parent.frame()) :
sd.est not defined for this family

[nirgrahamuk]

@Jem_Powell

glmmadmb

I am not at all familiar with this package, I believe its a non CRAN package ? therefore I don't know how to begin interpreting sd.est complaint.

Sorry