Figure 5.

Groups were significantly (a) more evenly distributed by sex but (b) more assorted by age class than expected from the randomized datasets. In these plots, the y-axis shows the ratio of the null models to the observed data (observed data are y=1). The phenotypic null model is shown by the red polygon and the spatio-temporally controlled null model is shown by the blue polygon. The sex ratio of groups up to a size of 10 is significantly closer to even (observed is more than random) than both the spatio-temporally restricted and the phenotypic randomizations (see electronic supplementary material, figure S6, for these shown as binomial probabilities). These groups also tend to contain more individuals that are either adults or juveniles than expected (groups are significantly less mixed than expected, observed is less than random). Plots show the ratio of the 95% confidence intervals of the randomizations to the observed data. Areas where the polygon overlap 1 are non-significant (the permutation is equal to the observed value). Plots are presented this way due to the difficultly of directly interpreting binomial probabilities, and the biological insignificance of the exponential decay as group size increases (see electronic supplementary material, figure S6).