Coefficient estimates () and standard errors (SE) for contributions to log-odds of significant explanatory variables (α = 0.05; based on Wald-statistics) of the multinomial logit regression models of habitat use in Turtle Mountain, Pembina Hills, and Porcupine Hills study areas, North Dakota, USA (data collected from 2016–17).

The 3 study areas were analyzed separately; land covers are the response categories that exhibited significantly different odds of use for the listed predictor level, relative to availability. Individual elk differed from other predictor variables in that only a subset was present in each herd; other predictors were tested in all study areas.