Table 1. Mixed-effect candidate models for the effect of noise playback on mass-dependent FCM concentrations (natural log-transformed).
| Modela,b | K c | ΔAICc d | wi e |
| Treatmentf | 5 | 0 | 0.66 |
| Treatment + Location | 6 | 2.4 | 0.20 |
| Treatment + Location + Treatment:Location | 7 | 4.7 | 0.06 |
| Null- random effects only | 4 | 5.5 | 0.04 |
| Treatment + Season | 6 | 6.5 | 0.03 |
| Treatment + Season + Treatment:Season | 7 | 10.0 | <0.01 |
| Treatment + NoiseType + Treatment:NoiseType | 7 | 10.8 | <0.01 |
| Treatment + Location + NoiseType + Treatment:Location + Treatment:NoiseType | 9 | 11.2 | <0.01 |
| Treatment + NoiseType + Season + Treatment:Season + Treatment:NoiseType | 9 | 20.7 | <0.01 |
| Treatment + MaxSize + Treatment:MaxSize | 7 | 25.3 | <0.01 |
| Treatment + NoiseType + Season + Treatment:NoiseType + Treatment:Season + Treatment:NoiseType:Season | 11 | 27.3 | <0.01 |
| Treatment + SpeakerDistance + Treatment:SpeakerDistance | 7 | 27.5 | <0.01 |
| Treatment + NoiseType + MaxSize + Treatment:NoiseType + Treatment:MaxSize | 10 | 35.4 | <0.01 |
| Treatment + NoiseType + SpeakerDistance + Treatment:NoiseType + Treatment:SpeakerDistance | 9 | 38.2 | <0.01 |
| Treatment + NoiseType + MaxSize + Treatment:NoiseType + Treatment:MaxSize + Treatment:NoiseType:MaxSize | 12 | 45.1 | <0.01 |
| Treatment + NoiseType + SpeakerDistance + Treatment:NoiseType + Treatment:SpeakerDistance + Treatment:NoiseType:SpeakerDistance | 11 | 60.4 | <0.01 |
a
Abbreviations of predictor variables in methods.
b
All models contain lek pairing and year as a random effect.
c
Number of parameters in the model.
d
Difference in AICc (Akaike's Information criteria for small sample size) values from the top ranking model.
e
Akaike weight (Probability that the model is the best fit model giving the data and model candidate set).
f
Model with substantial support (ΔAICc <2).