Eviews 进行异方差性检验及估计模型.docx

《Eviews 进行异方差性检验及估计模型.docx》由会员分享，可在线阅读，更多相关《Eviews 进行异方差性检验及估计模型.docx（9页珍藏版）》请在淘文阁 - 分享文档赚钱的网站上搜索。

1、异方差性检验及存在异方差模型估计检验使用方法：(1) G-Q检验(2) White检验模型估计方法：加权最小二乘法(WLS)下表为2000年中国局部省市城镇居民每个家庭平均年可支配收入(X)与消费性支出(Y)的统计数据:地区可支配收入X消费性支出Y地区可支配收入X消费性支出Y北京10349. 698493.49河北5661.164348.47天津8140.506121.04山西4724.113941.87内蒙古5129.053927. 75河南4766. 263830. 71辽宁5357.794356. 06湖北5524. 544644. 50吉林4810. 004020.87湖南6218.

2、735218.79黑龙江4912. 883824.44广东9761.578016.91上海11718. 018868.19陕西5124. 244276. 67江苏6800. 235323.18甘肃4916. 254126.47浙江9278.167020. 22山东6489.975022.00一、利用Eviews求出线性模型Dependent Variable: Y Method: Least Squares Date: 05/01/12 Time: 21:15 Sample: 1 20Included observations: 20CoefficientStd. Error t-Statis

3、ticProb.c272.2250159.64391.7052010.1054X0.7551520.02331132.394510.0000R-squared0.983137Mean dependent var5199.515Adjusted R-squared0.982200S.D. dependent var1625.275S.E. of regression216.8400Akaike info criterion13.69084Sum squared resid846352.2Schwarz criterion13,79041Log likelihood-134.9084Hannan-

4、Quinn criter.13,71027F-statistic1049.404Durbin-Watson stat1.189266Prob(F-statistic)0.000000可得模型：/ =272.225+ 0.755X,青海新疆5169. 965644. 864185. 734422.93(1.705)(32. 394) R2=0. 9832二、异方差检验(1)G-Q检验：首先将可支配收入X升序进行排列，然后去掉中间4个样本，将余下的样 本分为容量各为8的两个子样本，并分别进行回归。上海11718.018868.19青海5169.964185. 73北京10349.698493. 4

5、9内蒙古5129. 053927. 75广东9761.578016.91陕西5124. 244276. 67浙江9278.167020.22甘肃4916. 254126.47天津8140. 506121.04黑龙江4912. 883824.44江苏6800. 235323.18吉林4810. 004020. 87山东6489. 975022.00河南4766. 263830. 71湖南6218. 735218. 79山西4724.113941.87大样本大样本小样本样本取值较小的Eviews输出结果如下Dependent Variable: Y Method: Least Squares Da

6、te: 05/01/12 Time: 21:44 Sample: 1 8 Included observations: 8CoefficientStd. Errort-StatisticProb.c1277.1611540.6040.8290000.4388X0.5541260.3114321.7792870.1255R-squared0.345397Mean dependent var4016.814Adjusted R-squared0.236296S.D. dependentvar166.1712S.E. of regression145.2172Akaike info criterio

7、n13.00666Sum squared resid126528.3Schwarz criterion13.02652Log likelihood-50.02663Hannan-Quinn criter.12.87271F-statistic3.165861Durbin-Watson stat3.004532Prob(F-statistic)0.125501残差平方和：RSSf1 26528. 3样本取值较大的Ev i ews输出结果如下:Dependent Variable: Y Method: Least Squares Date: 05/01/12 Time: 21:49 Sample:

8、 1 8 Included observations: 8CoefficientStd. Error t-StatisticProb.C212.0683530.72240.3995840.7033X0.7619210.06033012.629290.0000R-squared0.963746Mean dependent var6760.477Adjusted R-squared0.957704S.D. dependent var1556.814S.E. of regression320.1754Akaike info criterion14.58793Sum squared resid6150

9、73.7Schwarz criterion14,60779Log likelihood-56.35173Hannan-Quinn criter.14.45398F-statistic159.4990Durbin-Watson stat1.722683Prob(F-statistic)0.000015残差平方和：RSS2=615073. 7口 Rss?一 因此统计量为：f = 7# = 48611在5%的显著性水平下，弓05（6,6） = 428,4864.28,因此拒绝原假设，存在异方差性。(2 ) White检验：在原模型的最小二乘估计窗口上选择“ ViewRes idua ITestsHe

10、teroskedasticity TestsWhite” 得到如下结果:Heteroskedasticity Test White检验统计量值为12.64768,查询x2005(2) = 5.99,因此12. 64785.99,因而拒绝原假设， 模型存在异方差。F-statistic14.62196Prob. F(2,17)0.0002Obs*R-squared12.64768Prob. Chi-Square(2)0.0018Scaled explained SS5.563250Prob. Chi-Square(2)0.0619三、估计存在异方差的经济模型利用加权最小二乘法(WLS)进行估计：

11、首先在对原模型进行估计后，保存残差，步 3聚如下：QuickGenerate Ser ies 再输入 “e1 =resid”，得到 e1QuickEstimte Equation 再输入 “Y C X”选择Options,在“Weighted LS/TLS”输入“1/a(e1)” (备注:abs表示绝对值) 得到如下结果；Dependent Variable: 丫Method: Least SquaresDate: 05/01/12 Time: 22:16Sample: 1 20Included observations: 20Weighting series: 1/ABS(E1)Coeffi

12、cientStd. Errort-StatisticProb.C415.4878116.98153.5517400.0023X0.7290590.02243032.504160.0000Weighted StatisticsR-squared0.983248Mean dependent var4471.622Adjusted R-squared0.982318S.D.dependent var7313.497S.E. of regression77.04359Akaike info criterion11.62126Sum squared resid106842.9Schwarz criter

13、ion11.72083Log likelihood-114.2126Hannan-Quinn criter.11.64070F-statistic1056.520Durbin-Watson stat1.622472Prob(F-statistic)0.000000Unweighted StatisticsR-squared0.981673Mean dependent var5199.515Adjusted R-squared0.980654S.D. dependentvar1625.275S.E. of regression226.0570Sum squared resid919831.6Durbin-Watson stat1.223571即采用加权最小二乘估计得到的回归方程:=415.48 + 0.7290Xj(3. 55)(32.50) R2=0. 98加权结果与不加权结果差异很大。