[R] Appropriate error distribution

[Cristina Gomes]

Hi,
My name is Cristina. I'm interested in studying which continuos predictor 
variables (such as grooming received, rank, etc.) affect grooming given, as 
well a continuos variable. I'm having problems finding an appropriate family 
distribution to fit the GLMM I'm doing. The response variable, grooming 
given, has many zeros which does not allow me to use a gamma distribution. I 
tried with a poisson but since I had to convert the data to integers I fear 
loosing a lot of information (and anyway my data is originally continuos and 
not counts). I found in the help archive that someone with a similar problem 
(continuos data and many zeros) was adviced to use Tweedie models. I don't 
know anything about this and wouldn't kow if this is appropriate in my case 
and possible to apply to generalized linear mixed models.
I'm wondering if anybody could provide me with any good insights on what 
distribution I could us and if, in the case of tweedie being a good option, 
if it can be used with GLMM's..
Thanks a lot,
Cristina.