, we then fitted another ordered beta regression but as a Generalized Linear Mixed Model (GLMM), also in the ordbetareg package (Kubinec 2022). To perform this analysis, we first defined each season based on the solstices and equinoxes in each year of the study; 2021 and 2022. We then specified frugivory as a function of the interaction between days elapsed within a season and the season of interest. We also included a random effect to account for differences among individuals and sites, as in the GAMM. The GLMM allowed us to quantify overall differences in fruit consumption between sites and seasons, while allowing us to estimate the rate of change in frugivory within seasons.
For all estimates, we calculated the Bayesian probability of direction (PD) from the posterior distribution (Makowski et al. 2019), which is as a measure of the probability that the effect is indeed in the observed direction (i.e., positive or negative). This was done by dividing the number of posterior draws in the direction of the posterior median estimate (i.e., negative or positive values indicating the directionality of the effect), by the total number of posterior draws.