《异方差性的检验和修正》由会员分享，可在线阅读，更多相关《异方差性的检验和修正（5页珍藏版）》请在金锄头文库上搜索。
1、计量经济学实验四一一异方差的检验和修正实验目的：学习建立回归模型，并进行异方差检验和对模型进行修正实验内容：1) 试用OLS建立居民人均消费支出和可支配收入的线性模型2) 检验模型是否存在异方差性3) 如果存在异方差性，试釆取适当的方法估计模型参数数据來源：第四章课后习题8下表列出了某年中国部分省市城镇居民家庭平均每个全年可支配收入(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浙江9279.167020.22大津8140. 506121.04ill东6489. 975022. 00河北5661.
2、 164348.47河南4766.263830. 71111西4724.113941.87湖北5524.544644.5内蒙占5129.053927. 75湖南6218. 735218.79辽宁5357. 794356. 06广东9761. 578016. 91吉林4810. 004020. 87陕西5124. 244276. 67黑龙江4912. 883824. 44甘肃4916. 254126.47上海11718.018868.19青海5169. 964185. 73江苏6800. 235323. 18新疆5644. 864422. 931、做Y关于X的散点图以及回归分析将数据通过exce
3、l录入到eviews中，对解释变量与被解释变量做散点图，选择解 释变量作为group打开，在数据表“ group冲点击view/giaph/scaMei/simple scatter,出现以上数据的散点图，如下图所示：在Eviews软件下，OLS (普通最小二乘法)估计结果如图所示:View I Proc IViewIProclObject| Print Name FreezeEstimate! Forecast! Stats I Resids IRange: 1Sample: 1X s cJlijreidDependent Variable: YMethod: Least SquaresDa
4、te: 07704/13 Time: 23:47Sample: 1 20Included observations: 20VariableCoefficientStd. Error VStatisticProb.C272.3635159.67731.7057130.1053X0.7551250.02331632.386900.0000R-squared0.983129Mean dependent回5199.515Adjusted R-squared0.982192S.D. dependentvar1625.275S.E. of regress io n216.8900Akaike info c
5、riterion13.69130Sum squared resid846743.0Schwarz criterion13.79087Log likelihood-134.9130Hannan-Quinn criter.13.71073F-statistic1048.912Durbin-Watson stat1.301684Prob(F-statistic)0.0000002、异方差的检验先采用G-Q检验。在对20个样本按X从人到小排序，去掉中间4个个体，对前后两个样本 进行OLS估计，样本容量为8。数据如卞：地区可支配收入X消费性支岀Y地区可支配收入X消费性支出Y上海11718.018868.
6、19青海5169.964185.73北京10349.698493.49内蒙古5129.053927.75广东9761.578016.91陕西5124.244276.67浙江9279.167020.22甘肃4916.254126.47天津8140.56121.04黑龙江4912.883824.44江苏6800.235323.18吉林48104020.87山东6489.975022河南4766.263830.71湖南6218.735218.79山西4724.113941.87前一个样本OLS估计结果如图Dependent Variable: YMethod: Least SquaresDate:
7、01/01/06 Time: 03:09Sample: 1 6Included observations: 8VariableCoefficient Std. Error t-Statistic Prob.C212.2118530.88920.3997290.7032X0.7618930.06034812.625050.0000R-squared0.963723Mean dependent var6760.477Adjusted R-squared0.957676S.D. dependent var1556.814S.E. of regression320.2790Akaike info cr
8、iterion14.58858Sum squared resid615472.0Schwarz criterion14.60844Log likelihood-56.35432F-statistic159.3919Durbin-Watson stat1.722960Prob(F-statistic)0.000015后一个样本OLS估计结果如图i Equation： IT1TITLEDlorkfile： 4-8 (2)1 ： :UntitledVView| Proc | Object | Print Name | Freeze| Eistimate Forecast) Stats | Resid
9、s |Depemderrt Variable: YMethod: Least SquaresDate: 01/OW6 Time: 03:03Sample: 1 8Included observations: 8VariableCo efficientStd. Error t-StatisticProb.c1277.1611540.6040.8290000.4388X0.5641260.3114321.7792870.1255R-squared0.345397Mean dependent var4016.814Adjusted R-squared0.236296S.D. dependent va
10、r166.1712S.E. of regression145.2172Akaike info criterion13.00666Sum squared resid126528.3Schwarz criterion13.02662Log likelihood-50.02663F-statistic3.165861Durbin-Walson stat3.004532Prob(F-stalistic)0.125501F=(RSSi/(8-1 -1 )/(RSS2/(8-1 -1 )=4.86在5%的显著水平下，自由度为(6,6)的F分布临界值Fo.o5 (6,6) =4.28,于是 拒绝无异方差性的
11、假设，表明原模型存在异方差性。3、估计模型参数首先，采用加权最小二乘法进行估计。在对原模型进行OLS估计后，在ewws的主菜单中 选择“quick/generate se门es.”在出现的对话框中输入“e=resid”，点击确定生成新数列e：为 了寻找适当的权，作In/关于x的ols回归，结果如下：Equation: UNTITLED Workfi.VanableCoefficientStd Errort-StatisticProb,C6 8251331 1264456 046266oooooX0.0004600.0001652 7922580.0120R-squared0 302237Mea
12、n dependent var9 827192Dependent Variable LOG(Ea2) Method Least Squares Date: 10/3009 Time: 09 46 Sample 1 20Included observations 201.53276942 26685 35 866681 8B2469S.D dependent var Akaike info cntenon Schwarz critenon F-statistic Prob(F-sttistic)Adjusted R-squared S E of regression Sum squared re
13、sid Log likelihood Durbin-Watson stat1 786002 3 796668 3 8062427 7967060.012035图的结果显示，X前的参数在5%的显著性水平卜不为零，同时，F检验也表明方程的线性 关系在5%的显著性水平下成立。其次，采用异方差稳健标准误法修正原OLS的标准差，得到下图所示的估计结呆：1 Equat ion:UNTITLED Workfi. 口巨*ViewjFroci!Objects.Print.Freeze !ForecastRends iDependent Variable YMethod: Least SquaresDate: 1
14、0/30&9 Time 1029Sample: 1 20Included observations 20White Heteroskedasticity-Consistent Standard Errors & CovarianceVanableCoefficientStd. Errorb StatisticProbc272 3635181 67571 4991740.1512X07551250.03163623 868860 00000 983129R-squared Adjusted R-squared S E. of regression Sum squared resid Log likelihood Durbin-Watson stat0 982192216846743.013491302 0879
