[R] GLM - guess the distribution of the response variable

[wonko at posteo.de]

Dear all,

I have question regarding GLMs:
I have a discrete response variable and a continuous explaining 
variable. Like this:
http://www.myimg.de/?img=example1db0f.jpg

I want to use a GLM to investigate. I have to specify the "familiy of 
the distribution of the response variable" - or, maybe more precise, the 
"family of the distribution of the residuals of the response variable" 
(the mean of the response should be explained by the deterministic model 
part of the GLM, as far as I understood).

How do I know the distribution family of the response? OK - its needs 
to be a discrete distribution, but there are Poisson, Binomial, Neg. 
Binomial.

Any ideas?

Many thanks,

wonko