Abstract
Compositional dietary data sets lend themselves to multivariate analyses, but differences in approaches used by authors worldwide hinder interpretations and comparisons of trophic relationships in different ecosystems. Dietary data are also affected by inconsistencies in sample numbers, the ingestion by each individual typically of a small subset of the range of food types and there are often differences in the taxonomic identification of food types within and between studies. Current approaches to multivariate dietary analyses are reviewed and categorised, and the best approaches to analyse dietary data detailed. These approaches encompass (1) determinations of the food type for analysis the dietary category, (2) the construction of units for analysis using randomisation - the dietary replicate representing sub-groups of those animals, rather than individuals, (3) recommendations for data transformation and choice of resemblance measure, (4) significance testing of the sometimes complex relationships between factors of interest, (5) linking of significant results to innovative and appropriate data visualisations, and finally to (6), the dietary category ingested by the animal. A pictorial summary of the entire approach is provided and these approaches can be easily applied using purchased or free-access software.
