Habitat selection of Israel desert rodents: comparison of a traditional and new method of analysis.

Studied habitat selection in six rodent species that occur in sandy or rocky areas of the Israeli desert using two complementary quantitative techniques. A distribution method for detecting habitat selection takes cognizance of density dependence but cannot establish what properties of the environment are being selected by the species. This and the traditional regression method were equally successful in discriminating between rodent species that exhibit selection and those that do not. Only the regression method can suggest the habitat variables that are actually preferred by a species. Either temporal or spatial variation in the data may result in backward regressions (indicating the right variable but in the wrong direction), because populations at low densities may be restricted to their best habitats. At high densities, however, populations utilize both preferred and marginal habitats. When both temporal and spatial variation exist, the regression method may fail altogether to detect habitat selection even when it exists. The distribution method does not suffer from this weakness.-from Authors