Factors affecting elephant speed between visits to waterholes.

Speed was regressed against explanatory variables in linear mixed models with elephant identity as a random effect on intercept. The 'Progression' variable was included as a quadratic predictor (i.e. the value and the square of the value ('Prediction2') was included in the model). The waterhole at the beginning of the trip is either different (commuting trip) or the same (looping trip) than the waterhole at the end of the trip. Progression in the trip was expressed as percent of the total trip duration. Estimates for the reference intercept (looping trips) and for the deviations associated to other levels of explanatory variables are presented, with 95% confidence intervals obtained by parametric bootstrap with 10000 samples. Estimates for which the 95% confidence interval do not include zero are in bold.