The discrete choice model was described as (McDonald et al. 2006):
where was an estimate of relative selection ratios for resource unit i; xij were habitat variables;
β1 – βp were coefficients of habitat variables; U’ were unique location indices; and A was the set of indices for available locations in the choice set. Based on the top model results, each location (i) in our dataset had a estimated relative selection ratio given by the above equation, where 1
= 2.516, 2 = 2.317, 3 = 1.618, 4 = 0.128, 5 = -0.486, 6 = 1.018, 7 = -0.956, 8 = 0.0033, 9 = - 0.000002, 10 = -0.0015, 11 = -0.0017, 12 = -0.0012, 13 = -0.0005, 14 = -0.0016, 15 = -0.0012, 16 = - 0.0012. Variables xi1-xi7 are 7 of 8 categorical landcover types, where corn field was the
reference with a -value of zero. Variable xi8 was the continuous distance to road measures and xi9-xi16 was the continuous distance to road measure interaction with each landcover type as described in Table 3.