BOOTSTRAPPING TO CREATE NON-PARAMETRIC CONFIDENCE INTERVALS FOR SELECTION RATIOS 
OF FEEDING SITES OF GREAT BLUE HERONS ON THE UPPER MISSISSIPPI RIVER SYSTEM.

Sarah Timm1, Christine Custer1,2, Douglas Olsen1.
1U.S. Geological Survey, Upper Midwest Environmental Science Center, 575 Lester 
Avenue, Onalaska, WI  54650, 2U.S. Geological Survey, Upper Midwest 
Environmental Science Center, 2630 Fanta Reed Road, La Crosse, WI  54603.

Selection ratios are often used in resource selection studies to give an 
indication as to whether or not a particular resource (often a habitat) is being 
selected, avoided, or used in proportion to availability.  Selection ratios are 
advantageous because they do not depend on what types of habitats are deemed 
available (unlike chi-squared tests).  Confidence intervals do, however, depend 
on the assumption that selection ratios are normally distributed, something that 
appears to be true only if each category (selected and available sites has a 
‘moderately large’ (e.g. n ? 5) sample size.  When these assumptions do not 
hold, bootstrapping can be used to create confidence intervals free of 
distributional assumptions.  Data from a study of Great Blue Heron (Ardea 
herodias) feeding sites in the Upper Mississippi River System are used as an 
example.  

Keywords:  Resource Selection, Selection Ratios, Bootstrap Confidence Intervals, 
(Ardea herodias)