Laserfiche WebLink
<br /> <br />efficient combination of predictors on which a classification <br />equation might be built. The nominal variables, land form and <br />vegetation zone, were coded according to the algebraic rules of <br />orthogonal coefficients (Myers and Well 1991:628) for three and <br />five levels, respectively, when subjected to discriminant analysis. <br />' The discriminant analysis was run on SPSS mainframe based software <br />on the University of Colorado VAX 8600 Network. Raw data used in <br />' the analysis are presented in Table 7. <br />' Results <br />Results of the independent samples t-tests are summarized in <br />' Table 10. Sites and nonsites differed significantly in relation to <br />slope and distance to water. On average, nonsite locations were <br />characterized by slopes 2-1l2 times greater than those of site <br />t locations and occurred 100 m (40~) farther from known water sources <br />than did site locations. <br />' t Results of the Chi-square analyses are presented in Table 8 <br />and Table 9. The observed frequencies of site and nonsite <br />locations within the three vegetation zones and across the three <br />land form categories differed by margins too great to be attributed <br />' to sampling error. Overall, sites tended to cluster in the <br />grassland and vegetation zone and on terraces, saddles, and ridges, <br />while nonsites occurred primarily in the shrub vegetation zone and <br />on hill slopes. <br />R total of four potential predictor variables were identified: <br />' slope, distance to water, vegetative cover, and land form. <br />Different linear combinations of these variables produced <br />t classifications functions of varying power and efficiency. Some <br />combinations were particularly good at assigning site locations to <br />' their true categories (true positives), but simultaneously <br />misclassified a large proportion of nonsite locations as sites <br />(false positives). Other combinations tended to categorize nonsite <br />' locations correctly (true negatives), produced relatively high <br />' 82 <br /> <br />