Abstract
Understanding how animals select resources provides important information regarding their requirements for survival. One method for studying diet selection is multiple-choice feeding preference experiments, in which animals are offered multiple food items simultaneously. However, because food items are offered simultaneously, the consumer's choices are not independent. They therefore violate assumptions of univariate statistical tests, such as ANOVAs and t-tests, which are most often used in such studies. I used multivariate statistics to examine diet selection of yellowbellied marmots (Marmota flaviventris) for the 7 plants most commonly found in their foraging areas. Using Hotelling's T2, I rejected the null hypothesis that the proportions of plant species consumed were equal to the expected value of 1/7 (T2 = 180.63, P = 0.025). Cafeteria-style feeding trials indicated that dandelions (Taraxacum officianale) were selected significantly more often than expected by a random distribution (P = 0.001), whereas grasses (Bromus spp. and Poa spp.) and cinquefoil (Potentilla gracilis) were avoided (P = 0.002). Use of the remaining species, cow parsnip (Heracleum lanatum), yarrow (Achillea millefolium), mountain blue bells (Mertensia fusiformis), and wild sweet pea (Lathyrus leucanthus), did not differ significantly from random, although cow parsnip and wild sweet pea tended to be more preferred and avoided, respectively, than the remaining species. Preferences in these marmots may have been based on fatty acid content, which is important for hibernation.