When to use offset in binomial glm

Question

I am trying to model the number of successes in data where the number of trials is not fixed. I am trying to fit a binomial generalised mixed effects model, with the number of trials as an 'offset' variable, like so:

        glmer(cbind(number of successes, number of failures) ~ 
         offset(number of trials) + other fixed and random 
         effects, data = data, family = binomial())

Is this the correct way to use offset?

Why not a Poisson with an offset? – dimitriy Dec 06 '21 at 23:51 — dimitriy, Dec 06 '21 at 23:51

score 1 · Answer 1 · answered Dec 06 '21 at 23:22

1

The information on the number of trials is already there. The number of trials = number of successes + number of failures. You can provide inputs for logistic regression as a column of zeroes and ones, probabilities, or two columns for successes and failures. In neither of the cases, there is a reason to provide the number of trials as a separate variable.

answered Dec 06 '21 at 23:22

Tim

138,066

That's helpful clarification, thanks. – Ellen Dec 14 '21 at 11:17

When to use offset in binomial glm

1 Answers1