Post-flight initiation distance response and correlation analysis
Given that squirrels could only take one of two escape options, we first
ran a binomial model exploring the factors that influenced whether a
squirrel sheltered or not (Table S2) and then we ran separate bivariate
models to analyze these disparate post-FID responses and their
among-individual correlations with FID. Specifically, one model
contained FID and shelter emergence time as response variables, and the
second model contained FID and stop look distance as response variables.
Shelter emergence time and stop look distance were both modelled using a
Gamma distribution and log-link function. Both bivariate models
contained the same set of predictors (year, sex, life stage, trial
number, home site activity score and trappability). Walker identity and
squirrel identity were included as random intercepts. To estimate the
correlation between an individual’s FID and post-FID response, we
extracted the posterior mean among-individual random intercept
correlation from each model respectively. We also estimated the
repeatability (R) of each post-FID response by extracting the variance
components from their respective models and using the following formula
for Gamma distributed variables (Nakagawa et al. 2017):
\begin{equation}
R=\ \frac{V_{\text{ind}}}{V_{\text{ind}}+V_{\text{obs}}+ln(1+\frac{1}{\upsilon})}\nonumber \\
\end{equation}where Vind is the among-individual variance,
Vobs is the among-observer variance, and \(\upsilon\) is
the shape parameter of the gamma distribution.