FLOOD09697 - Laserfiche WebLink

Home Browse Search

(7) (8) (9) + .0205X6X14 + .400X7X15 + .150X8X16 . . . (5.21) Standardized form: (I) (2) (3) Y = -.159XI+.0476X2X9+.0181X2(1/XIO) (4) (5) (6) + .421X3Xl1 + .212X4X12. .0168X5XI3 (7) (8) (9) +.0367~X14 + .704X7X15 +.262X8XI6 . . . . (5.22) The r2 is .827, only slightly more (.004) than without terms # 7 and # 9. The variables are defined in Table 5.2, and the terms are identified in Table 5.3. The only coefficients significant at the .05 level are those for terms I and 8; those for terms 4 and 5 are significant at a level between .10 and .20. Elim- inating variables until all those remaining are signifi- cant al the .05 level results in Equation 5.16. The added terms in Equation 5.21 add little to Ihe pre- dictIve abillty. Term 9, a Type IV term for the influence of public opinion, was the lhird mosl important term as judged by the coefficients of the standardized form (5.22). This term may be more significant than indi- cated in the previous paragraph because of the high multicollinearity of the equation. High muiticollinear. ity means thai much of the variance that would be explained by one term could be accounted for by other terms in Ihe equation and consequenlly the sizes of the coefficients are smaller (Johnston, 1972; Theil, 1970) and unstable (Gordon, 1968). Therefore, the values of the coefficients of this equation should not be relied on, but the output of the equation could still be relatively stable because fluctuations in some coefficients would be compensated by opposite fluclu- ations in others. The results of using Equation 5.21 in the model appear reasonable. The influence of lhe previous de. cision agency evaluation dominates, and the sign of Term 9 representing the influence of the public's evalu- ation is in the direction expected. A favorable public evaluation reinforces the rating by the decision agen- cy. This direct relationship may be offset some by the smaller value of the regression constant in Equa. tion VI (5.21) compared to Equation IV (5.16), mean- ing that the other terms would have to be more posi- tive to have the same output value of the equation than in Equation IV. The relative influence of public and previous agency evaluation terms is more than apparent from the coefficients of Equation N (5.16). This is be- cause the public and previous agency's evaluations are weighled by variables X7 and X8. When the val- ues of lhe decision agency as listed in Table 5.4 are substituted in Equation VI, the following equation is derived: Y = -.0466- .21OX1 + .0170X9 + L362/XIO + .00612Xl1 + 210X12 - .140XI3 + .0333X14 + .800X15 + .570X16 . . . . . (5.23) The coefficients of XIS and XI6 are much more equal in value in Equation 5.23 than are those of the corres- ponding terms # 8 and #9 in Equation 5.21. XIS and XI6 are the outputs of other equations in the mod- el. For Equation 5.21, XIS is input from Equation 5.16 applied to the decision agency. X 16 comes from the public evaluation equation discussed below. Population Evaluation Equation Equation V(a): Population evaluation equation - continuous form Unstandardized form: (I) (2) (3) (4) POPEVE = .645 - .129Xl + 129X2 - .491X3 +.047X4 (5) (6) (7) + .060X5 + .024X6X12 + .033X7XI3 (8) (9) + .0085X8X14 + .0081X9X15 (10) (II) + .0090XIOX16 + .0062Xl1 XI7 (5.24) Standardized form: (j) (2) (3) (4) POPEVE = .645-.177X1 +.140X2-.1IOX3+.105X4 (5) (6) (7) + .084X5 + .345X6X12 + .202X7XI3 (8) (9) + .105X8X14 + .128X9X15 (10) (II) + .087XIOX16 + .1l3XllX17 (5.25) The variables in Equations 5.24 and 5.25 are identified in Table 5.6. The terms are listed in Table 5.9. This equation has an r2 of .492 for the sample and flood control proposals used. The significance levels of the terms (denoted by bracketed numbers) are: (I) .005; (2) .003; (3) .015; (4) .02; (5) .07; (6) .0001; (7) .006; (8) .0627; (9) .02; (10) .05; and (II) .02. All the terms in Equation 5.25 were reasonably significant, and the r2 was fair by social science stand- 71