# glm link function HELP! To gaussian, or not to gaussian, that is the question...

#### Rose1226

##### New Member
Please ignore question no longer need help - thank you for all of your responses!

Last edited:

#### duskstar

##### New Member
Ignore this post I made, poisson is used for rates whilst binomial is used for proportions
If I'm understanding you correctly your outcome is a rate, and if that is the case then you do need to use a Poisson and (I think) canonical link

Last edited:

#### Dason

##### Ambassador to the humans
I just used the default family=gaussian. My variable I want to model is visitation rate, which is recorded in percent (0-100%). If I had a feeding station out for 14 days and it was used by small mammals 7 out of 14 days than it had a visitation rate of 50%.
I don't understand why you want to use a poisson here. That doesn't make sense to me at all. Do you have the original data for # of visits and # of days? If you have that then why not use a binomial family?

#### Dason

##### Ambassador to the humans
You totally can use a binomial family on data that actually is... binomial
example code:
Code:
set.seed(500)
len <- 50
x <- seq(0,1, length.out = len)
B1 <- .6
B0 <- .2
p <- exp(B1*x + B0)/(1 + exp(B1*x + B0))
n <- rbinom(len, 20, .7)
y <- rbinom(len, n, p)
o <- glm(cbind(y, n-y) ~ x, family = binomial)
summary(o)
I generated some fake data and then analyzed it.