成都市05年度住宅市场定价模型Word格式文档下载.docx

1、占地面积x5室x6厅x7卫厨卫装修标准精装修初装修清水房x8周边环境有大型商场,超市累加有大学、小学、中学、幼儿园交通便利与否运动设施x9户型多层7层及其以下,无电梯小高层1.5y高层7层以上的电梯公寓（一）模型的建立：首先我们假定模型为线性形式:即使用eviews对该模型进行回归分析（原始数据参见附件一）Dependent Variable: YMethod: Least SquaresDate: 07/06/05 Time: 15:49Sample: 1 46Included observations: 46VariableCoefficientStd. Errort-StatisticP

2、rob. X1330.3166158.94512.0781810.0449X235.7124018.361931.9449150.0596X3270.5117200.50991.3491190.1857X4-0.1459410.618733-0.2358710.8149X5-75.59086209.5936-0.3606540.7205X6-201.6988314.9935-0.6403270.5260X7406.5982350.68531.1594390.2539X8555.8820129.37934.2965310.0001X9636.7978429.48781.4826910.1469C

3、-2332.0101320.955-1.7653960.0860R-squared0.783368 Mean dependent var3184.696Adjusted R-squared0.729209 S.D. dependent var1440.707S.E. of regression749.7082 Akaike info criterion16.26691Sum squared resid20234248 Schwarz criterion16.66444Log likelihood-364.1388 F-statistic14.46446Durbin-Watson stat1.0

4、97759 Prob（F-statistic）0.000000回归结果见上表，发现t统计量的p值明显大于0.05，无法通过检验，修正R2=0.729209较高，F值显著大于临界值0.05，我们认定模型中存在多重共线；与此同时，DW的值1.097759在0.01显著性水平下落入无法判定区域（0.927 1.834）。鉴于回归结果不太理想，我们决定将模型进行对数变化。对数变换后模型中参数可以看作Y相对于X的百分比变化，可以降低数据中因为评分的人为因素造成的不利影响，而且还可以减少异方差的影响。即原表达式变化为：根据新的对数模型，我们采取逐步回归的方式不断的减少解释变量个数寻求最佳的log（y）表达

5、式，即当模型中有i（i9、8、7.）个解释变量时，我们从个模型中选取F统计量的p值最优的一个，直至达到理想效果时（即F统计量的p值小于给定的临界值）停止减少变量。我们将这个F的p值取为0.05，应用统计软件SAS，我们可以直接得到逐步回归的最优结果，如下：Step 1 当引入九个解释变量时的回归结果： ErrorSum of SquaresFProbc4.870083880.861217541.5197691231.98LOGX10.571785640.089300261.9484601941.00LOGX20.747594780.232611170.4909099610.330.0028LO

6、GX30.006137600.155745480.000073810.000.9688LOGX40.038679100.043761930.037127030.780.3826LOGX50.044340810.144569850.004470770.090.7608LOGX6-0.103417070.190909930.013946260.290.5914LOGX7-0.045819030.159050780.003944120.080.7749LOGX80.062661310.129374750.011148860.230.6311LOGX90.119580310.173489300.022

7、578980.480.4951显然，LOGX3、LOGX4LOGX9的回归结果皆不尽如人意，其F的p值大于0.05。该模型达不到要求。因此，我们改用8个变量回归，并在个回归结果中选出最优的一个（即下表）。有必要说明，这一选择过程由SAS软件完成，依据的标准已事先在程序内部设定。同理，以下7步皆照此处理，不再赘述。Step 2 引入8个解释变量的回归结果SumofSquares4.879325000.817419621.6477045935.630.573466450.077390362.5391731954.910.747036700.229025560.4920009510.640.0024

8、0.037857210.037949830.046017981.000.32500.044570360.142490090.004524520.100.7562-0.100720910.175808330.015177850.330.5702-0.046032750.156798870.003985640.77070.062770690.127587840.011192970.240.62560.116094800.147222730.028755820.620.4354Step 3 引入7个解释变量的回归结果4.973617890.742581912.0245787544.860.56744

9、5550.073720742.6739146759.250.722978070.211274940.5284852011.710.00150.036304480.037124860.043158790.960.33430.017714620.107927390.001215850.030.8705-0.101225970.173673260.015331920.340.56340.066887400.125280850.012864640.59650.114774510.145373900.028131740.4347Step 4 引入6个解释变量的回归结果4.979670320.732354

10、972.0345260946.230.569597770.071634492.7822591663.230.721163610.208336840.5272794211.980.00130.038203050.034834000.052929131.200.2795-0.086142090.145516500.015420970.350.55730.061697460.119703010.011690410.270.60920.122058090.136697530.035084650.800.3774Step 5 引入5个解释变量的回归结果coefficient4.951270820.723544812.0228363446.83Logx10.598675830.043737258.09354515187.36Logx20.746880040.200408810.5999656013.890.0006Logx40.038812080.0