Predictors of Fighting Ability
Two male traits were significant predictors of fighting
ability in the final model: the height of the casque and the
relative area occupied by the pink patch in the centre of
the green flank (Table 3). The estimated effect sizes of casque
height and flank patch, as measured by standardized
regression coefficients (Table 3), are roughly the same.
For example, if males A and B are separated by two standard
deviations in casque height, with, say, A greater
than B, and all else equal, the probability that A defeats
B is estimated to be exp(2 0.51)/[1 þ exp(2 0.51)]¼
0.74. For males differing in flank patch area, the corresponding
probability is 0.72. Of the ‘experience’ formulations,
the number of wins in the previous two fights
explained the most variability and was a strong predictor
of fighting ability (Table 3). Note that the coefficient
estimate of this variable should not be interpreted as indicative
of the strength of an ‘experience’ effect per se because
it may be confounded with aspects of male fighting
ability that we have not measured. The final model correctly
predicted the outcome of 87 of the 107 contests
(81%), suggesting that the model has good predictive
capacity. Neither mass nor SVL was a significant predictor
of fighting ability in the B–T model. Of the 63 contests
involving males that differed by more than 10% mass or
SVL, 29 (46%) were won by the larger male and 34
(54%) by the smaller male, a ratio that does not differ significantly
from that expected by chance (binomial test:
P ¼ 0.61).
Predictors of Fighting Ability
Two male traits were significant predictors of fighting
ability in the final model: the height of the casque and the
relative area occupied by the pink patch in the centre of
the green flank (Table 3). The estimated effect sizes of casque
height and flank patch, as measured by standardized
regression coefficients (Table 3), are roughly the same.
For example, if males A and B are separated by two standard
deviations in casque height, with, say, A greater
than B, and all else equal, the probability that A defeats
B is estimated to be exp(2 0.51)/[1 þ exp(2 0.51)]¼
0.74. For males differing in flank patch area, the corresponding
probability is 0.72. Of the ‘experience’ formulations,
the number of wins in the previous two fights
explained the most variability and was a strong predictor
of fighting ability (Table 3). Note that the coefficient
estimate of this variable should not be interpreted as indicative
of the strength of an ‘experience’ effect per se because
it may be confounded with aspects of male fighting
ability that we have not measured. The final model correctly
predicted the outcome of 87 of the 107 contests
(81%), suggesting that the model has good predictive
capacity. Neither mass nor SVL was a significant predictor
of fighting ability in the B–T model. Of the 63 contests
involving males that differed by more than 10% mass or
SVL, 29 (46%) were won by the larger male and 34
(54%) by the smaller male, a ratio that does not differ significantly
from that expected by chance (binomial test:
P ¼ 0.61).
การแปล กรุณารอสักครู่..
