Laserfiche WebLink
1 <br />key hydrological variables (e.g., mean particle size, elevation, timing and magnitude of peak <br />flow) predicted riparian plant distributions reasonably well. <br />The seven variables included in our multivariate analysis are presented in Table 3.2. <br />Since these factors are all related to fluvial geomorphic processes, it seems likely that not all <br />variables are necessary for predicting species distributions. The importance of our seven <br />explanatory factors was evaluated using discriminant function analysis. This statistical technique <br />is similar to multivariate regression, but it uses explanatory variables that can be continuous or <br />categorical to predict a categorical response variable. The response variable in our analysis was <br />the presence or absence of giant whitetop. The analysis included a sample of 53 random points <br />within Ouray National Wildlife Refuge. Input values were recorded from existing maps. The <br />analysis was performed using SAS's procedure for stepwise discriminant function analysis, <br />which iteratively eliminates variables with little influence on the response (i.e. presence or <br />absence of giant whitetop) and maximizes the rZ value. The accuracy of the model's predictions <br />was assessed using a jackknife approach on the original 53 observations. <br />Lam' <br /> <br />r <br /> <br /> <br /> <br />11 <br /> <br />w? <br />24