Figure 1
Schematic representation of the modelling approach used in this study to estimate the consistent individual-level differences in emergence behaviour. Observed emergence depended on a combination of a binomial and continuous process which we represented as a hurdle model: i.e. whether the individual emerged or not during the trial (binomial), and if it emerged, how long it took to leave the tube (continuous). Each of these processes was modelled as an intercept (average across all experimental trials) and a slope (change in probability or time with experience from each additional experimental trial). Because individuals were tested multiple times, these intercepts and slopes were modelled at a lower hierarchical level using the identity of each individual to estimate consistent differences between individuals. The Bayesian framework allowed us to examine these differences between individuals in two ways: (1) by extracting the posterior distributions that described the mean and range of the variation at the individual level, and (2) by directly estimating the mean for each individual’s ‘personality score’ for each of the model parameters (i.e. behavioural traits) of interest.
