计量经济学实验报告 .doc

《计量经济学实验报告 .doc》由会员分享，可在线阅读，更多相关《计量经济学实验报告 .doc（18页珍藏版）》请在淘文阁 - 分享文档赚钱的网站上搜索。

1、2 1978-2011年的数据搜集年份人均GDP城市化率城镇居民家庭人均可支配收入政府支出城镇居民消费水平197838117.92343.41122.09405197941918.964051281.79425198046319.39477.61228.83489198149220.16500.41138.41521198252821.13535.31229.98536198358321.62564.61409.52558198469523.01652.11701.02618198585823.71739.12004.25765198696324.52900.92204.91872198711

2、1225.321002.12262.189981988136625.811180.22491.2113111989151926.211373.92823.7814661990164426.411510.23083.5915961991189326.941700.63386.6218401992231127.462026.63742.222621993299827.992577.44642.329241994404428.513496.25792.6238521995504629.0442836823.7249311996584630.484838.97937.5555321997642031.

3、915160.39233.5658231998679633.355425.110798.1861091999715934.78585413187.6764052000785836.22628015886.568502001862237.666859.618902.5871132002939839.097702.822053.15738720031054240.538472.224649.95790120041233641.769421.628486.89867920051418542.991049333930.28941020061650043.911759.540422.7310423200

4、72016944.9413785.849781.351190420082370845.6815780.862592.661352620092557546.5917174.776299.9315025 3 REVIEWS模型建立及检验3.1 散点图变化分析3.1.1 GDPP（人均GDP）和CSH（城市化）的关系3.1.2 GDPP（人均GDP）和JMKZPSR（城镇居民家庭人均可支配收入）的关系3.1.3 GDPP（人均GDP）和ZFZC（政府支出）的关系3.1.4 GDPP(人均GDP)和GMXFSP（城镇居民消费水平）3.2 Ganger检验3.2.1首先，我们研究GDPP和CSH的因果检

5、验。Pairwise Granger Causality TestsDate: 06/03/12 Time: 10:42Sample: 1978 2009Lags: 1Null Hypothesis:ObsF-StatisticProb.CSH does not Granger Cause GDPPP310.782470.3839GDPPP does not Granger Cause CSH0.571930.4558由表可知，CSH影响GDPP的概率较大，故可以将CSH作为自变量，GDPP为因变量。3.2.2其次，我们研究GDPP和JMKZPSR的因果检验。Pairwise Granger

6、Causality TestsDate: 06/03/12 Time: 10:44Sample: 1978 2009Lags: 1Null Hypothesis:ObsF-StatisticProb.JMKZPSR does not Granger Cause GDPP310.248210.6222GDPP does not Granger Cause JMKZPSR0.194840.6623由表可知， JMKZPSR影响GDPP的概率高，故可以将JMKZPSR作为自变量，GDPP作为因变量。3.2.3紧接着，我们研究GDPP和ZFZC之间的因果关系。Pairwise Granger Caus

7、ality TestsDate: 06/03/12 Time: 10:45Sample: 1978 2009Lags: 1Null Hypothesis:ObsF-StatisticProb.ZFZC does not Granger Cause GDPP310.020240.8879GDPP does not Granger Cause ZFZC0.337200.5661由表可知，GDPP和ZFZC相互影响，概率都比较大，所以可以将ZFZC作为自变量。3.2.4最后，我们研究GDPP和GMXFSP的因果关系。Pairwise Granger Causality TestsDate: 06/0

8、3/12 Time: 10:44Sample: 1978 2009Lags: 1Null Hypothesis:ObsF-StatisticProb.JMXFSP does not Granger Cause GDPP3016.02510.0004GDPP does not Granger Cause JMXFSP7.442160.0111由表可知，GDPP和 JMXFSP的相关可能性都非常低，顾将JMXFSP作为自变量剔除。3.3选择模型形式，做回归,描绘模型估计模型： Dependent Variable: GDPPMethod: Least SquaresDate: 06/07/12 T

9、ime: 16:47Sample: 1978 2011Included observations: 34VariableCoefficientStd. Errort-StatisticProb.C472.7725178.03882.0.0126CSH2-1.0.-3.0.0006ZFZC0.0.8.0.0000JMKZPSR1.0.14.663990.0000R-squared0.Mean dependent var7863.882Adjusted R-squared0.S.D. dependent var9292.254S.E. of regression250.9664Akaike inf

10、o criterion13.99865Sum squared resid.Schwarz criterion14.17822Log likelihood-233.9770Hannan-Quinn criter.14.05989F-statistic15070.08Durbin-Watson stat1.Prob(F-statistic)0.令 3.4随机误差项的正态性检验（JB检验） 通过JB检验发现，估计模型随机误差项可能为正太分布的可能性P5%,所以通过检验。3.5 Ramsey reset test检验Ramsey RESET Test:F-statistic4.Prob. F(1,27

11、)0.0533Log likelihood ratio4.Prob. Chi-Square(1)0.0337Test Equation:Dependent Variable: GDPPMethod: Least SquaresDate: 06/03/12 Time: 13:59Sample: 1978 2009Included observations: 32VariableCoefficientStd. Errort-StatisticProb.C-44.45361313.7799-0.0.8884CSH2-0.0.-0.0.7963JMKZPSR1.0.13.922750.0000ZFZC

12、-0.0.-0.0.9270FITTED28.81E-064.36E-062.0.0533R-squared0.Mean dependent var6325.906Adjusted R-squared0.S.D. dependent var7066.021S.E. of regression246.1018Akaike info criterion13.99197Sum squared resid.Schwarz criterion14.22099Log likelihood-218.8715Hannan-Quinn criter.14.06788F-statistic6382.086Durb

13、in-Watson stat1.Prob(F-statistic)0.Prob.F值为0.5335%,所以模型被误设可能性较小。3.6 T、F检验，拟合优度检验t-Statistic2.-3.13.981707.T值的绝对值2，通过检验，说明此模型拟合优度较好。Prob(F-statistic)0.F值为0，远远小于5%，说明此模型拟合优度较好。R-squared0.=0.99，说明改模型可行性很大，拟合度好。3.7 Wald Test检验，若Prob. F5%，接受约束条件Wald Test:Equation: UntitledTest StatisticValuedfProbability

14、F-statistic3.(1, 28)0.0749Chi-square3.10.0644Null Hypothesis Summary:Normalized Restriction (= 0)ValueStd. Err.-1 + C(2)2 - 3*C(3) + C(4)2.1.Delta method computed using analytic derivatives.3.8邹氏突变检验：若Prob. F5%，认为该点很可能是突变点通过观察整体数据较为平稳，未发现明显突变点，其中对1995年、2004年进行随机检测，如下图：Chow Breakpoint Test: 1994Null

15、Hypothesis: No breaks at specified breakpointsVarying regressors: All equation variablesEquation Sample: 1978 2009F-statistic10.66037Prob. F(4,24)0.0000Log likelihood ratio32.68074Prob. Chi-Square(4)0.0000Wald Statistic42.64146Prob. Chi-Square(4)0.0000Chow Breakpoint Test: 2004Null Hypothesis: No br

16、eaks at specified breakpointsVarying regressors: All equation variablesEquation Sample: 1978 2009F-statistic51.32985Prob. F(4,24)0.0000Log likelihood ratio72.22598Prob. Chi-Square(4)0.0000Wald Statistic205.3194Prob. Chi-Square(4)0.0000所以通过邹氏检验，发现无突变点。3.9模型的比较：观察AIC和SC值的变化，若有下降的现象，该模型可能更好些。Dependent

17、Variable: GDPPMethod: Least SquaresDate: 06/07/12 Time: 19:12Sample: 1978 2009Included observations: 32VariableCoefficientStd. Errort-StatisticProb.C-355.7275157.9942-2.0.0327CSH21.0.2.0.0141ZFZC-0.0.-4.0.0001JMKZPSR1.0.17.919050.0000JMKZPSR25.91E-057.81E-067.0.0000R-squared0.Mean dependent var6325.

18、906Adjusted R-squared0.S.D. dependent var7066.021S.E. of regression149.3804Akaike info criterion12.99347Sum squared resid.2Schwarz criterion13.22249Log likelihood-202.8955Hannan-Quinn criter.13.06938F-statistic17333.87Durbin-Watson stat1.Prob(F-statistic)0.此时原模型通过比较发现 增加一个变量后的模型更适合4 REVIEWS异方差检验及克服4

19、.1异方差检验4.1.1图形法4.1.2 WHITE检验Heteroskedasticity Test: WhiteF-statistic4.Prob. F(9,22)0.0023Obs*R-squared20.53007Prob. Chi-Square(9)0.0149Scaled explained SS25.76099Prob. Chi-Square(9)0.0022Test Equation:Dependent Variable: RESID2Method: Least SquaresDate: 06/03/12 Time: 14:01Sample: 1978 2009Included

20、 observations: 32VariableCoefficientStd. Errort-StatisticProb.C8464.488.70.0.9798CSH2-201.35391601.322-0.0.9011(CSH2)20.1.0.0.8457(CSH2)*JMKZPSR-0.0.-0.0.7542(CSH2)*ZFZC0.0.0.0.9304JMKZPSR-13.46487270.5561-0.0.9608JMKZPSR20.0.0.0.6731JMKZPSR*ZFZC-0.0.-0.0.7512ZFZC46.1234650.200820.0.3682ZFZC20.0.0.0

21、.8448R-squared0.Mean dependent var58835.95Adjusted R-squared0.S.D. dependent var.6S.E. of regression76913.92Akaike info criterion25.58907Sum squared resid1.30E+11Schwarz criterion26.04711Log likelihood-399.4251Hannan-Quinn criter.25.74090F-statistic4.Durbin-Watson stat2.Prob(F-statistic)0.，由white检验知

22、，在，查分布表，得临界值，所以拒绝原假设，接受备择假设，表明模型存在异方差。4.2异方差的修正Dependent Variable: GDPPMethod: Least SquaresDate: 06/03/12 Time: 14:34Sample: 1978 2009Included observations: 32Weighting series: 1/RESID2VariableCoefficientStd. Errort-StatisticProb.C849.1712171.22644.0.0000CSH-48.907558.-5.0.0000ZFZC0.0.17.868100.000

23、0JMKZPSR1.0.41.039550.0000Weighted StatisticsR-squared0.Mean dependent var633.4496Adjusted R-squared0.S.D. dependent var3246.371S.E. of regression1.Akaike info criterion3.Sum squared resid36.13133Schwarz criterion3.Log likelihood-47.34883Hannan-Quinn criter.3.F-statistic.8Durbin-Watson stat1.Prob(F-

24、statistic)0.Unweighted StatisticsR-squared0.Mean dependent var6325.906Adjusted R-squared0.S.D. dependent var7066.021S.E. of regression309.4658Sum squared resid.Durbin-Watson stat0.4.3再次对修正后的模型做white检验Heteroskedasticity Test: WhiteF-statistic1.09E+22Prob. F(1,30)0.0000Obs*R-squared32.00000Prob. Chi-S

25、quare(1)0.0000Scaled explained SS3.01E-10Prob. Chi-Square(1)1.0000Test Equation:Dependent Variable: WGT_RESID2Method: Least SquaresDate: 06/03/12 Time: 14:41Sample: 1978 2009Included observations: 32Collinear test regressors dropped from specificationVariableCoefficientStd. Errort-StatisticProb.C2.4

26、5E-165.44E-174.0.0001WGT23.14E-083.00E-191.05E+110.0000R-squared1.Mean dependent var1.00E-06Adjusted R-squared1.S.D. dependent var5.68E-06S.E. of regression3.03E-16Sum squared resid2.75E-30F-statistic1.09E+22Durbin-Watson stat1.Prob(F-statistic)0.,所以修正后的模型通过WHITE检验得到无异方差。此时模型为： 5 REVIEWS自相关检验及克服5.1

27、自相关检验5.1.1 DW检验法Dependent Variable: GDPPMethod: Least SquaresDate: 06/07/12 Time: 17:28Sample: 1978 2009Included observations: 32VariableCoefficientStd. Errort-StatisticProb.C459.4286200.79842.0.0299CSH2-1.0.-3.0.0021ZFZC0.0.7.0.0000JMKZPSR1.0.13.981700.0000R-squared0.Mean dependent var6325.906Adjus

28、ted R-squared0.S.D. dependent var7066.021S.E. of regression259.3089Akaike info criterion14.07039Sum squared resid.Schwarz criterion14.25360Log likelihood-221.1262Hannan-Quinn criter.14.13112F-statistic7663.496Durbin-Watson stat1.Prob(F-statistic)0. 说明在滞后一期时 该模型存在一阶自相关5.1.2 LM检验法 Breusch-Godfrey Seri

29、al Correlation LM Test:F-statistic11.63691Prob. F(1,27)0.0021Obs*R-squared9.Prob. Chi-Square(1)0.0019Test Equation:Dependent Variable: RESIDMethod: Least SquaresDate: 06/07/12 Time: 18:03Sample: 1978 2009Included observations: 32Presample missing value lagged residuals set to zero.VariableCoefficien

30、tStd. Errort-StatisticProb.C-164.8782177.6396-0.0.3615CSH20.0.0.0.4776ZFZC-0.0.-1.0.1990JMKZPSR0.0.0.0.8403RESID(-1)0.0.3.0.0021R-squared0.Mean dependent var-4.19E-13Adjusted R-squared0.S.D. dependent var246.4424S.E. of regression220.7472Akaike info criterion13.77451Sum squared resid.Schwarz criteri

31、on14.00354Log likelihood-215.3922Hannan-Quinn criter.13.85043F-statistic2.Durbin-Watson stat1.Prob(F-statistic)0. LM检验也说明该模型在滞后一期时存在一阶自相关。5.2 广义差分法克服自相关滞后一期时，两边同时乘以 并将原模型与所得模型相减得到差方后模型： 6 REVIEWS多重共线检验及克服6.1多重共线检验GDPPCSH2ZFZCJMKZPSRGDPP10.470610.181010.73518CSH20.4706110.089080.06614ZFZC0.181010.089

32、0810.96026JMKZPSR0.735180.066140.9602616.1.1 去掉后 对模型重新进行计算Dependent Variable: GDPPMethod: Least SquaresDate: 06/07/12 Time: 19:02Sample: 1978 2009Included observations: 32VariableCoefficientStd. Errort-StatisticProb.C-167.625290.48260-1.0.0741ZFZC0.0.9.0.0000JMKZPSR0.0.20.686300.0000R-squared0.Mean

33、dependent var6325.906Adjusted R-squared0.S.D. dependent var7066.021S.E. of regression302.4890Akaike info criterion14.35103Sum squared resid.Schwarz criterion14.48844Log likelihood-226.6164Hannan-Quinn criter.14.39657F-statistic8443.399Durbin-Watson stat0.Prob(F-statistic)0.此时 所以不应该被剔除6.1.2 去掉ZFZC后 对

34、模型重新进行计算Dependent Variable: GDPPMethod: Least SquaresDate: 06/07/12 Time: 19:05Sample: 1978 2009Included observations: 32VariableCoefficientStd. Errort-StatisticProb.C995.1454327.71263.0.0050CSH2-3.0.-5.0.0000JMKZPSR1.0.23.808460.0000R-squared0.Mean dependent var6325.906Adjusted R-squared0.S.D. dependent var7066.021S.E. of regression450.9395Akaike info criterion15.14960Sum squared resid.Schwarz criterion15.28702Log likelihood-239.3937Hannan-Quinn criter.15.19515F-statistic3791.291Durbin-Watson stat0.Prob(F-statistic)0.此时 所以ZFZC不应该被剔除6.1.3 去掉JMKZPSR后