Laserfiche WebLink
Appendix A. Examples of models developed to test for the ability of combinatio <br />habitat variables to discriminate between backwaters used by Color <br />and those not used by Colorado squawfish in the Green River within <br />Utah: 1990-1996. Models were developed using stepwise discriminan <br />and tested for accuracy of classification using discriminant funct <br />Significance level for entry into models was set at pa0.15 and pri <br />for classification were proportionate to group frequency. <br />POOLED SAMPLESa <br />Stepwise selection: <br /> Average <br />Step Variable R2 Wilks~ (Squared ~ <br /> Lambda Canonical u <br /> Correlation ,I <br />1 mean width of backwater 0.1102 0.8898 0.1102 ~~ <br />2 mean depth of backwater 0.0030 0.8872 0.1128 (~ <br />3 <br />l ~ surface area of backwater 0.00094 0.8863 0.1137 <br />Classification error rate: <br /> No Yes Total ~ <br />Rate 0.0858 0.3929 0.1766 <br />Priors 0.7044 0.2956 <br />a Model for pooled samples was developed by comparison of all other models <br />judgement of the researchers and random testing rather than through step <br />discr-i~inant analysis. It is designed to be simple and as effective as <br />for easy application in the field. <br />