ABSTRACTAccurately assessing primate diets is important in studies of behavioral ecology and evolution. While previous research has compared sampling methods (scan, focal), we examined how sampling schedule influences accuracy of dietary measures. We define sampling schedule as the combined distribution (random vs. consecutive) and frequency of sampling days within a given month. Under field conditions, time may be required to locate a study group, and we therefore also subtracted 1, 2, or 3 h from the beginning of all non‐consecutive days in each sampling schedule to mimic observation time lost to search. From a dense (near daily) 5‐year record of feeding behavior derived from focal animal follows of adult females in five wild blue monkey (Cercopithecus mitis) groups, we created data subsets matching various sampling schedules, and compared monthly dietary measures calculated from each subset to those based on the full data set. These measures included (1) the proportion of observation time feeding on fruit, (2) diet composition (three top‐ranked food items), (3) species richness of plant diet, (4) Shannon–Wiener diversity index based on plant species, and (5) Holmes–Pitelka index expressing dietary overlap with the previous month. We used generalized linear mixed models to assess how frequency and sampling type (a combination of distribution and hours lost) relate to a subset's deviation from the full data set, where a smaller deviation (or higher chance of matching, for diet composition) implies greater accuracy. For all dietary measures, increasing observation frequency increased accuracy. The response to distribution varied among measures, but sampling types generally differed more at lower frequencies. Deviation varied widely within and between dietary measures, and some sampling schedules resulted in values with large percentage differences from the “full” data. Accordingly, when designing and comparing studies, researchers should consider how sampling schedules may influence the accuracy of the dietary measures of interest.
Read full abstract