Data from: An updated global dataset for diet preferences in terrestrial mammals: testing the validity of extrapolation

1. Diet is a key trait of an organism’s life history that influences a
broad spectrum of ecological and evolutionary processes. Kissling et al.
(2014) compiled a species-specific dataset of diet preferences of mammals
for 38% of a total of 5364 terrestrial mammalian species assessed for the
International Union for Conservation of Nature’s Red List, to facilitate
future studies. The authors imputed dietary data for the remaining 62% by
using extrapolation from phylogenetic relatives. 2. We collected dietary
information for 1261 mammalian species for which data were extrapolated by
Kissling et al. (2014), in order to evaluate the success with which such
extrapolation can predict true diets. 3. The extrapolation method devised
by Kissling et al. (2014) performed well for broad dietary categories
(consumers of plants and animals). However, the method performed
inconsistently, and sometimes poorly, for finer dietary categories,
varying in accuracy in both dietary categories and mammalian orders. 4.
The results of the extrapolation performance serve as a cautionary tale.
Given the large variation in extrapolation performance, we recommend a
more conservative approach for inferring mammalian diets, whereby dietary
extrapolation is implemented only when there is a high degree of
phylogenetic conservatism for dietary traits. Phylogenetic comparative
methods can be used to detect and measure phylogenetic signal in diet. If
data for species are needed, then only the broadest feeding categories
should be used. This would ensure a greater level of accuracy and provide
a more robust dataset for further ecological and evolutionary analysis.