Skip to main content

Advertisement

Table 1 Forward selection of 30 landscape variables

From: Habitat selection by American beaver at multiple spatial scales

Model DIC
Intercept 663.95
cropdist 661.06
shrubbd 661.60
shrubfq 661.62
cropdist + shrubfq 658.96
shrubbd + cropdist 659.24
shrubfq + hwbd 659.59
cropdist + shrubbd + wwetbd 657.21
cropdist + shrubbd + wwetbd + hbwetfq 655.15
  1. Step 1 selected single variables, which reduced the deviance information criterion by > 2.0, to form single-variable models. Step 2 added one of the remaining 29 variables to each of three selected single-variable models, respectively, resulting in three two-variable models which reduced DIC by > 2.0. Step 3 added 28 remaining variables to each of three resulting two-variable models, resulting in a three-variable model. Step 4 added each of remaining 27 variables to the selected three-variable model and resulted in a four-variable model. All 26 five-variable models had DIC values greater than that of the selected four-variable model