Data from: Inferring predator-prey interactions in food webs

1. Food webs are a powerful way to represent the diversity, structure, and
function of ecological systems. However, the accurate description of food
webs requires significant effort in time and resources, limiting their
widespread use in ecological studies. Newly published methods allow for
the inference of feeding interactions using proxy variables. Here, we
compare the accuracy of two recently described methods, as well as
describe a composite model of the two, for the inference of feeding
interactions using a large, well-described dataset. 2. Both niche and
neutral processes are involved in determining whether or not two species
will form a feeding link in communities. Three different models for
determining niche constraints of feeding interactions are compared, and
all three models are extended by incorporating neutral processes, based on
relative abundances. The three models compared here infer niche processes
through 1) phylogenetic relationships, 2) local species trait
distributions (e.g. body size), and 3) a composite of phylogeny and local
traits. 3. We show that all three methods perform well at predicting
individual species interactions, and that these individual predictions
scale up to the network level, resulting in food-web structure of inferred
networks being similar to their empirical counterparts. 4. Our results
indicate that inferring food-web structure using phylogenies can be an
efficient way of getting summary webs with minimal data, and offers a
conservative test of changes in food-web structure, particularly when
there is low species turnover between sites. Inferences made using traits
requires more data, but allows for greater understanding of the mechanisms
underlying trophic interactions. A composite model of the two methods
provides a framework for investigating the importance of how phylogeny,
trait distributions, and relative abundances, affect species interactions,
and network structure.