【正文】
ATRIX procedure:****** Ridge Regression with k = ******Mult R .802353780RSquare .643771588Adj RSqu .611387187SE ANOVA table df SS MSRegress Residual F value Sig F .00001172Variables in the Equation B SE(B) Beta B/SE(B)x1 .025805860 .003933689 .574462395 x4 .004531316 .007867533 .050434658 .575951348Constant .357087614 .741566536 .000000000 .481531456 END MATRIX Y對(duì)x1 x2 x3 做嶺回歸Run MATRIX procedure:****** Ridge Regression with k = ******Mult R .850373821RSquare .723135635Adj RSqu .683583583SE ANOVA table df SS MSRegress Residual F value Sig F .00000456Variables in the Equation B SE(B) Beta B/SE(B)x1 .016739073 .003359156 .372627316 x2 .156806656 .047550034 .275213878 x3 .067110931 .032703990 .159221005 Constant .754456246 .000000000 END MATRIX 由圖及表可知,(1)y 與x1 x2 x3 x4 ,.(2) y對(duì)其余四個(gè)變量的線性回歸方程為 由于的系數(shù)為負(fù),說(shuō)明存在共線性,固所得的回歸系數(shù)是不合理的。(3) 由于條件數(shù)=10,說(shuō)明存在較強(qiáng)的共線性。(4) 由上表可知由后退法和逐步回歸法所得到的線性回歸方程為 由于的系數(shù)為負(fù),說(shuō)明仍然存在共線性。(5) Y對(duì)其余四個(gè)自變量的嶺回歸如上表所示。(6) 選取嶺參數(shù)k=,得嶺回歸方程,回歸系數(shù)都能有合理的解釋。(7) 用y對(duì)x1 x2 x3 做嶺回歸,選取嶺參數(shù)k=,嶺回歸方程為回歸系數(shù)都能有合理的解釋,由 B / SE(B) 得近似的t值可知,x1 x2 x3 都是顯著的,所以y對(duì)x1 x2 x3的嶺回歸是可行的。