Laserfiche WebLink
<br /> <br />sites could be used reliably in analysis of the six environmental <br />variables outlined in the previous section. A pilot study using <br />' this sample attempted to determine whether a valid predictive model <br />could be constructed for the project area. Since this undertaking <br />was preliminary and limited in scope, a purely empirical design was <br />' used. However, variables extracted from ethnographic analogy and <br />informant interview may be incorporated at a later date as <br />' appropriate. <br />' Method <br />As noted in the previous section, there was no effect upon the <br />' six environmental variables of interest for site location on Indian <br />Mountain versus Rabbit Mountain. Therefore, a sample of 22 <br />randomly selected nonsite points within current project areas on <br />Indian Mountain and the western margin of Dowe Flats was compared <br />with the overall site sample (n=22) to determine which <br />' environmental variables discriminated between site and nonsite <br />locations. <br />' Two categories of data were used. Distance to water, <br />elevation, aspect, and slope were continuous metric variables that <br />' occurred on a ratio scale. Accordingly, differences between <br />average site and nonsite locations for each of these variables were <br />assessed by independent samples t-test (two-tailed). The <br />' variables, vegetation and land form, were discrete and occurred on <br />nominal scales. Differences bet:aeen the frequencies of site and <br />' nonsite locations across these variables were assessed by Chi- <br />square analysis. In order to satisfy the underlying assumptions of <br />' the Chi-square test, three types of land form, saddles, ridges and <br />hill tops were treated as a single category in the analysis. If <br />these categories were not combined, power of the test would have <br />been jeopardized by low cell frequencies. Those variables which <br />differed significantly across site and nonsite locations were <br />' manipulated via direct discriminant analysis to find the most <br />' B1 <br /> <br />