Statistical procedures

To study the variation of diets attributed to different covariates, we employed a permutational multivariate analysis of variance [45] using the 'adonis' function from 'vegan' R package [46]. In addition, the multivariate homogeneity of group dispersions (variances) was tested with the function 'betadisper' from 'vegan' package, a multivariate analogue of Levene's test for homogeneity of variances.

We estimated and compared diet diversities using the approach proposed by Pallmann et al. [47]. Instead describing diet diversity through a given index as, for example, Simpson or Shannon indices, we converted these "raw" indices into "true" diversities which all belong to one and the same mathematical family. That is, regarding different measures as special cases of Hill's general definition of diversity measures [48]. In this way, to study differences in diversity between the diet and availability, or among diets of different years, months, seasons or sites, we performed two-tailed tests for integral Hill numbers of orders -1≤q≤3. This selection includes the transformed versions of the three following indices: the species richness index, H_sr (q = 0), the Shannon entropy index, H_sh (q→1) and the Simpson concentration index, H_si (q = 2) [49]. All comparisons among diversities of groups were made with Tukey-like contrasts employing a resampling procedure. We did 5000 bootstrap replications so as to obtain reliable p-values [50]. Methods described here are implemented in R package “simboot” [51] and are fully described in Pallmann et al. [47]. All calculations were done in R version 3.0.3 [52].