I need to use GLM for count data. I'm thinking Possion GLM. How close mean and the variance be, to say they are equal?
If you are worried about overdispersion you can switch to the negative binomial distribution. I usually switch because the biological and ecological data I deal with is generally overdispersed. Subjectively, I switch for anything greater than 1.2, but for me I am generally dealing with values of 2 to 3.