Laserfiche WebLink
<br />The stream measures were grouped into "packages" according to each variable's <br /> <br />correlation with the others. The packages and correlations are shown in Table I. A <br /> <br />correlation coefficient of one indicated that the two variables were always carried out in the <br /> <br />restoration projects together and therefore were exactly the same in the regression, and in <br /> <br />fact influence each other. For example, because both "acquire land" and "establish an <br /> <br /> <br />education trail" variables were needed together, the correlation between those measures is <br /> <br />. <br /> <br />one. <br /> <br />Table I about here <br /> <br />Based on the review of the literature and the partial R2's, a subset of the variables in <br /> <br />each package was selected as candidate variables which were most likely to determine the <br /> <br /> <br />value owners place on a property. Omitting a stream restoration variable that is correlated . <br /> <br />with an included restoration variable may lead to mis-specification and bias in the included <br /> <br /> <br />variables. We recognize this problem and consider the estimated coefficient on the included <br /> <br />stream restoration variable to reflect the joint effect of all the stream restoration variables <br /> <br /> <br />that it is correlated with in its package. The robustness of using different stream restoration <br /> <br /> <br />variables from the same package in the hedonic regression is then tested by comparing the <br /> <br /> <br />resulting sign, size and statistical significance of the coefficients in Table II and the resulting <br /> <br /> <br />marginal values in Table m. <br /> <br />Table IT about here <br /> <br />. <br /> <br />10 <br />