Use of Multiple Regression and Use-Availability Analyses in Determining Habitat Selection by Gray Squirrels (Sciurus Carolinensis)
Multiple regression and use-availability analyses are two methods for examining habitat selection. Use-availability analysis is commonly used to evaluate macrohabitat selection whereas multiple regression analysis can be used to determine microhabitat selection. We compared these techniques using behavioral observations (n = 5534) and telemetry locations (n = 2089) of gray squirrels (Sciurus Carolinensis) on the Piedmont National Wildlife Refuge (PNWR) in Georgia. Use-availability analysis of stands classified according to their composition of pine and hardwood basal area produced inconsistent results; no pattern of selection was evident because similarly classified stands (e.g.. pine/ hardwood) received differing levels of use. In multiple regression analysis, tree species that predicted relative use by gray squirrels differed by season. Deciduous holly (Ilex decidua), sweetgum (Licpidambar styraciflua), water oak (Quercus nigra), willow oak (Q. Phellos), winged elm (Ulmus alata), and yellow-poplar (Liriodendron tulipifera) explained the most variation in seasonal stand use by gray squirrels; none of the 17 structural variables measured contributed significantly to predictive models. We found poor to moderate concordance (14.3 - 71.4%) between stand use predicted from multiple regression analysis and stand use determined by use-availability analysis. Our findings suggest that examinations of selection at different scales may result in differing interpretations of habitat use and erroneous inferences regarding habitat selection.