Table 1. Set of generalized linear mixed effect models (Restricted Estimate of Maximum Likelihood) with different random structures and different measures of elk age, either allowing individuals to change behaviour between years or not.
Elk have been monitored for multiple years, and the terminology ‘true age’ implies the actual age of the elk in a given year. The term ‘age at capture’ implies the age of the elk kept constant to that recorded at the beginning of the monitored period. ‘True age’ allows models to account for behavioural adjustments with age (learning), while ‘age at capture’ does not allow depicting learning processes. The 5 a priori models were run to explain the variability of three different response variables (log step-length, use of terrain ruggedness, use of forest). The top ranked structure (#5) selected using AIC was the same for all response variables. Because model selection was performed on models with different random effect structures, we opted to use the number of levels of the random effects minus 1 as the punishment for added random effects when calculating the AIC.
| # | Random intercept for elk identity (ID) and random slope for true ageX | Random intercept for year of study | Elk age estimate included in the model | ΔAICa [response variable: log step-length] |
ΔAICb [response variable: use of terrain ruggedness] |
ΔAICc [response variable: use of forest (0 = no forest, 1 = forest)] |
Model details | Model key word |
|---|---|---|---|---|---|---|---|---|
| 1 | None | (1|year) | True age | 6075.6 | 28966.2 | 11248.7 | No random effect for individual elk, age allowed to vary | Learning and selection at work, no individuality |
| 2 | None | (1|year) | Age at capture | 6160.9 | 28966.8 | 11160.6 | No random effect for individual elk, age not allowed to vary | Only selection at work, no individuality |
| 3 | (1|ID) | (1|year) | True age | 2880.8 | 2184.7 | 4147.8 | Animals can change behaviour (learning) between years, but they all learn in the same way (same slope). | Learning and selection at work |
| 4 | (1|ID) | (1|year) | Age at capture | 2882.1 | 2115.1 | 4124.6 | Animals cannot change their behaviour (no learning) between years. | Only selection at work |
| 5 | (True Age|ID) | (1|year) | True age | 0 | 0 | 0 | Individual animals can change behaviour (learning) as they age | Both individual learning and selection at work |
XA model with age at capture as random slope is not within the alternative models as such age metric does not change over time.
aFixed effects in the model: month + canopy cover + canopy cover^2 + terrain ruggedness + terrain ruggedness^2 + age*day of the week + age*time of the day + age*distance to road + age * hunting season.
bFixed effects in the model: month + canopy cover + canopy cover^2 + age*day of the week + age*time of the day + age*distance to road + age * hunting season
cFixed effects in the model: month + terrain ruggedness + terrain ruggedness^2 + age*day of the week + age*time of the day + age*distance to road + age * hunting season