I'm working with recruitment data for caribou.
There are 13 different herds, sampled over 20+ years, once per year. Some herds are sampled consistently, some only a few times over 20 years.
I'm testing recruitment to see if it is influenced by several different variables. i.e. Is recruitment affected if there is more harvesting of alternate prey? Is recruitment affected by more disturbance on their range? etc.
I'm using mixed effects models to account for variation in sampling effort between herds.
I'm wondering: do I include the HERD and the YEAR as random factors to account for differences in sampling effort, and to account for differences between years of samples?
Recruitment is a measurement of number of animal calves to number of cows. It is taken yearly by helicopter surveys where you count animals with children and without. This tells you how many are 'recruited' into the next years population. The other variables are measurements of what may affect where they live, and how that may affect how many calves are recruited into the next years population. There are 13 herds that have recruitment data. But the recruitment surveys vary in numbers per herd, and among years.
Ultimately I'm using linear mixed modelling (can be run as GLMM as well if recruitment is turned into a binomial distribution, which it can be).
Here is my model structure: M1z<-glmer(RECRUITMENT~ zNT_0 + (1|YEAR) + (1|HERD), family=gaussian, data=M_data_Z)
Here is a sample of the data. Recruitment takes a value between 0 and 100. There are about 220 different 'samples' unevenly distributed between caribou herds. Note: I standardize the data before I run the model so the effect sizes are similar for each predictor.