I'm considering a mixed-effects model to try to understand factors that influence the number of ticks sampled on wild rodents. My data is nested so that I have one tick count per rodent, multiple rodents per site and multiple sites per year (sites are repeated across years but not every site is present in all years).
So far, without fixed effects, my model looks something like this:
glmer(Ticks~(1|Year)+(1|Site), family=poisson, data=tickdata)
I understand that this will account for random variation between sites and between years. My main concern is that I think the effect of specific sites will change between years, e.g. some sites may recieve more rainfall than others, but the identity of the sites with the highest rainfall differs between years.
So do I need some sort of interaction term? Maybe something like this:
glmer(Ticks~(1|Year)+(1|Site)+(1|Year:Site), family=poisson, data=tickdata)