Generalized linear model (using a logit link) of vigilance probability in mouflon females, Caroux-Espinouse massif, France.

SEs were not meaningful here because the fitted probabilities were extremely close to zero ([60]; pgs. 197–198).
The analysis of deviance table (i.e., difference of deviances between successive nested models) gives the effects of age, quality of feeding sites, presence of horns, group size, lambing season, group composition, average visibility in the home range, hunting, area and 2 two-ways interactions on vigilance probability (see Table 1; Full model: Age+Repro compF+Lambing+Feeding+Horn+Hunting×Area+Visibility×Group size). Variables within brackets were evaluated concurrently to the preceding related variable (see Methods for details). Parameter values with its standard error are given for the best model (significant terms in bold). DF, degrees of freedom and SE, standard error.