Bo Blommendaal,

A few questions come to mind: what type of response is "Aggression" (i.e., integer scale, continuous, nominal)? Can the effects of zoo and individual be separated—in other words how do you know it's a zoo effect and not just the individual? Is there enough data points enough to make inferences to greater population?

I might be inclined to model it as Aggression ~ Group + Sex, random = ~1|ID, where ID is the individual (this may be what Focal is...).

Caleb A. Aldridge, thanks for your response!

"Aggression" is the rate per hour measured for each individual every year (the amount a behavior is performed divided by the total observation time in hours for that year). So I guess the response is continuous (?).

I think effects of zoo and individual cannot be seperated, no..

I have 11 individuals and for every individual four data points, so no, I do not think I can make inferences to greater population?

Focal is indeed the individual in my code.

I can share my dataset here with you if that helps to give you an idea!

Is the rate an average of multiple observations for each individual in each year? I.e., is the rate fractional or integer? The reason I ask is if the response is integer the Poisson distribution may be a good error term.

Caleb A. Aldridge

Yes the rate is indeed an average, does that make it integer?

No, integers are discrete, i.e. 1, 2, 3... perhaps fit the model without the zoo term and inspect the residuals for violations of assumptions, plot fitted v. observed for linearity, and you may be interested is caterpillar plot of random variable (gorilla I'd).