We review compositional analysis for testing whether animals use habitats in proportion to their availability (i.e., random use). We illustrate aspects of this method with goshawk habitat use data collected in southeast Alaska. Compositional analysis addresses several inadequacies of other habitat selection testing approaches. As the appropriate summary measure for use with compositional analyses, we propose the geometric mean selection ratio. However, caution must be used in interpreting results based on compositional analysis. It is not clear what definition of selection (i.e., analysis method) is best in all cases, and hypothesis tests of selection allow only narrow conclusions to be drawn without additional data or assumptions.

Key Words
Accipiter gentilis; Compositional analysis; Geographic Information System; Habitat use; Northern goshawk.

Grey W. Pendleton, Kimberly Titus, and Richard E. Lowell are Biometrician, Regional Supervisor, and Wildlife Biologist, respectively, Alaska Department of Fish and Game, Division of Wildlife Conservation, P.O. Box 240020, Douglas, AK, 99824. Eugene DeGayner is Wildlife Biologist, USDA Forest Service, P.O. Box 309, Petersburg, AK, 99833. Craig J. Flatten is Wildlife Biologist, Alaska Department of Fish and Game, Division of Wildlife Conservation, 2030 Sealevel Drive, Suite 205, Ketchikan, AK, 99901.