Laserfiche WebLink
<br />1 <br />' 6.2 Predictive Model of Prehistoric Site Locations <br />Introduction <br />' Predictive models attempt to generate reliable estimates of <br />archaeological site frequencies and locations in specific <br />geographic areas. The assumption underlying all such models is <br />' that a significant amount of the variance in site frequency can be <br />captured by the covariance of certain environmental traits. Site <br />' presence or absence is a dependent variable and environmental <br />traits are independent variables or predictors. Modeling attempts <br />' to find the dimension or dimensions along which site and nonsite <br />locations differ and generate a classification equation to predict <br />' group membership. Predictor variables usually are identified by <br />analysis of a sample of site and nonsite locations drawn from an <br />intensively surveyed area that is physiographically similar to, and <br />preferably contiguous with, the region of interest. <br />Predictive models serve both practical and theoretical <br />' purposes. From a practical standpoint, an effective predictive <br />model streamlines and economizes the compliance process by <br />' facilitating informed land use decisions. Identification of high <br />and low probability areas for archaeological sites allows <br />developers and land use planners greater flexibility in the initial <br />stages of the planning process. Managers may plan their projects <br />around anticipated results rather than reacting to discoveries at <br />' a later stage. From a theoretical perspective, predictive models <br />constitute extremely powerful tools for hypothesis testing and <br />theory building. Regardless of their intended use, technical <br />constraints on predictive models include the following: <br />' 1) Site and nonsite locations must differ significantly in <br />relation to certain environmental variables. <br />' 2) The number of predictor variables that can be combined to <br />create a prediction equation is constrained by the number of site <br />and nonsite locations available for analysis. <br />79 <br />1 <br />