Statistical approaches to measuring speciesŐ associations in mixed-species bird flocks
Emily Silverman, US Fish & Wildlife Service, Maryland
Measuring species associations in flocks is important for
both a theoretical understanding of flock organization and the design of
conservation plans for bird communities where flocks are common. Traditional
analyses of species associations use tables of co-occurrence, followed by
Chi-Square tests. Problems with this approach include a potential lack of
independence between flocks and an inability to determine species associations
for species that are abundant in flocks.
In addition, when multiple habitats are sampled, positive and negative
patterns may be the result of habitat affinities, and the spatial structure of
the data is ignored. We develop a model to generate flock data in which species
associations are known. We compare
and contrast traditional analyses of association with a new approach that uses
locations, and not flocks, as the unit of analysis. This approach measures the density of birds inside and
outside of flocks and the number of flocks by location, thus accounting for the
differences between locations in the pool of birds available to flock and the
opportunity to enter flocks.