. 2018 Jan 29;6(1):coy001. doi: 10.1093/conphys/coy001

Table 2:

Candidate models evaluated to predict the age class (immature or adult) of 86 male brown bears and 83 female brown bears. For each model, with the exception of the null model, we refined an initial model that contained all variables of interest to a final model by backward elimination of variables. Each final candidate model was selected by comparing ΔAICc values, i.e. ΔAICc = 0.00

Model	Initial structure	Refining procedure
1. Null	Intercept only	None
2. Hormone^a only (no interactions)	test + test² + test³ + prog + prog² + prog³ + est + est² + est³ + cort + cort² + cort³	Backward elimination
3. Hormone only (with interactions)	Most supported model 2 + all two-way interactions	Backward elimination
4. Hormone + time	Most supported model 3 + all hormone × day interactions + all hormone × year interactions	Backward elimination
5. Hormone + contact method	Most supported model 3 + all hormone × contact method interactions	Backward elimination
6. Hormone + study area	Most supported model 3 + all hormone × study area interactions	Backward elimination

^aHormones are testosterone (test), progesterone (prog), estradiol (est) and cortisol (cort).