Laserfiche WebLink
<br />. <br /> <br />. <br /> <br />. <br /> <br />To correctly account for changes in stream attributes, properties which were sold <br /> <br /> <br />before the restoration projects commenced were coded with zeroes for the restoration <br /> <br /> <br />measures. Of the seven funded projects in the sample, all but one were initiated in the <br /> <br /> <br />summer of 1989. On average, restoration projects were completed in one year and a half. <br /> <br />Estimating the Hedonic Equation <br /> <br />As shown in equation (1) a hedonic equation is specified as a function of structural, <br /> <br />neighborhood, and environmental variables. We chose variables that represented each of the <br /> <br />three categories, In our data set, many of these variables within each category were <br /> <br /> <br />correlated with each other. Therefore, the next step in variable selection is to conduct an <br /> <br /> <br />analysis for multicollinearity among the candidate explanatory variables. A correlation matrix <br /> <br /> <br />was calculated using Econometric Views (Lilien et al. 1994). Many of the stream <br />characteristic variables are highly correlated with each other, having correlation coefficients <br />greater than 0.80. In addition, many of the property and demographic characteristics are <br />highly correlated. To avoid high sampling variances and low t-statistics, variables must be <br />chosen to minimize the effect of multicollinearity. <br />The second step undertaken was to conduct regression analyses on groups of <br />independent variables to calculate partial RI.s. Results from this test should indicate which <br />variables in each of the three groups (property, stream and demographic characteristics) are <br />most influenced by the others. Independent variables with low partial RI,s within the three <br />groups of projects are preferred to minimize multicollinearity. <br /> <br />9 <br />