I have a simple question regarding using time as a random effect.

Briefly, I am modelling population-level changes in the movement rates of 68 woodland caribou using generalized additive mixed models that have been specified using the function gamm4 from the R package gamm4. I have calculated the daily movement rate for each animal over a two year period (i.e., km/day per animal). Since repeated rates were calculated for each caribou, I am including the animal id (AID) as a random intercept in my GAMM. An example of my model is shown below:

model1<-gamm4(daily.rate~s(yrday, bs="cc", k=-1), random=~(1|AID), data=caribou)

where daily rate is actually the log of the raw daily rate, yrday is the day of the year (1-365) and AID is individual animal id (001....068).

However, I also want to account for potential annual variation in movement rates. As I understand it, year and animal id are crossed random effects which means that if I were to include Year (1 or 2) as a random effect, I would specify the model as follows:

    model2<-gamm4(daily.rate~s(yrday, bs="cc", k=-1), random=~(1|AID)+(1|Year), data=caribou)

BUT, I also understand that random effects should have at least 5-6 levels, otherwise it's difficult to measure among-level variance. Since I only have 2 levels to the variable Year (1 and 2), I don't think I should be including it as a random effect. My question is this: should I include it as a fixed effect? Any advice would be much appreciated!


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.