where 𝑥i is the number of seeds of dispersal syndrome i in the sample of interest, for all n dispersal syndromes. This transformation resulted in a compositional dataset, ideal for performing Correspondence Analysis (CA), which we conducted using the easyCODA package (Greenacre 2019).
CA is a multivariate technique that ordinates compositional data, allowing to visualize the association between grouping variables and the various parts of the composition (Greenacre 2017). In our case, the grouping variable is the frugivory level. We ordinated the normalized seed counts in the CA, plotted the 99% confidence intervals for each frugivory level (No, Low, and High), and followed the ordination with a permutational multivariate analysis of variance (PERMANOVA) using the vegan package (Oksanen et al. 2022). This was done to test for differences in the seed dispersal syndromes dispersed based on the level of frugivory exhibited. The PERMANOVA was performed on the original count data, given its suitability for analyzing ecological count data.
We performed the PERMANOVA with 10,000 permutations based on Bray-Curtis dissimilarity. Additionally, we created a distance matrix using the Bray-Curtis method and performed a multivariate homogeneity of group dispersions analysis (PERMDISP) to assess dispersion differences between frugivory levels. An ANOVA was conducted on the PERMDISP object, followed by a Tukey post-hoc test to determine which frugivory levels differed significantly in the seed syndromes dispersed.