您好,欢迎来到意榕旅游网。
搜索
您的当前位置:首页异方差例题Word版

异方差例题Word版

来源:意榕旅游网


第五章习题答案演示

一、数据如下:(表5.1) Y X

2 8777 105 9210 90 99 131 10508 122 10979 107 11912 406 12747 503 13499 431 14269 588 15522 8 16730 950 17663 779 18575 819 19635 1222 21163 1702 22880 1578 24127 16 25604 1400 26500 1829 26760 2200 28300 2017 27430 2105 29560 1600 28150 2250 32100 2420 32500 2570 35250 1720 33500 1900 36000 2100 36200 2800 28200

二、数据输入EVIEWS软件,注意输入过程中要定义e2

1 / 11

“quick”菜单下“estimate equation”结果如下:(表5.2)

Dependent Variable: Y Method: Least Squares Date: 04/13/08 Time: 16:01 Sample: 1 31

Included observations: 31 Variable C X R-squared

Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat Coefficient -700.4110 0.087831 Std. Error 116.6679 0.004827 t-Statistic -6.003458 18.19575 Prob. 0.0000 0.0000 846.7570 13.790 13.99042 331.0852 0.000000 0.9194 Mean dependent var 1266.452 0.916686 S.D. dependent var 244.4088 Akaike info criterion 1732334. Schwarz criterion -213.4175 F-statistic 1.0829 Prob(F-statistic) 最小二乘估计结果如下: Estimation Command:

===================== LS Y C X

Estimation Equation:

===================== Y = C(1) + C(2)*X

Substituted Coefficients: =====================

Y = -700.4109607 + 0.08783115594*X

三、检验模型的异方差:

(一)图形法

1、EViews软件操作。

由路径:Quick/Qstimate Equation,进入Equation Specification窗口,键入“y c x”,确认并“ok”,得样本回归估计结果,见表5.2。

(1) 生成残差平方序列。在得到表5.2估计结果后,回到以下界面:

点击“procs”下的“generate series”菜单,输入公式“e2=(resid)^2”

回到以下界面(e2已经生成,即残差序列生成):

(2) 在该界面下:单击“views”菜单下的“multple graphs”下的“scatter”, 操

作如下:

(3) 显示结果如下(图5.1):

3000

2000Y1000001000020000X4000003000040000300000E2200000100000001000020000X3000040000

2ei(4)从图5.3分析可知:大致看出残差平方随Xi的变动呈增大的趋势,因此,模

型很可能存在异方差。但是否确实存在异方差还应通过更进一步的检验。

(二)Goldfeld-Quanadt检验 1、EViews软件操作。

(1)对变量取值排序(按递增或递减)。在Procs菜单里选Sort Series命令,出现排序对话框,如果以递增型排序,选Ascenging,如果以递减型排序,则应选Descending,键入X,点ok。本例选递增型排序,这时变量Y与X将以X按递增型排序。

(2)构造子样本区间,建立回归模型。在本例中,样本容量n=31,删除中间1/4的观测值,即大约9个观测值,余下部分平分得两个样本区间:1—11和21—31,它们的样本个数均是11个,即n1n211。

由路径:Quick/Qstimate Equation,进入Equation Specification窗口,键入“y c x”,确认并“ok”,得样本回归估计结果,(注意:sample:1 11)

(表5.3)

Dependent Variable: Y Method: Least Squares Date: 04/13/08 Time: 22:26 Sample: 1 11

Included observations: 11 Variable C X R-squared

Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat Coefficient -744.6351 0.088258 Std. Error 195.4108 0.015705 t-Statistic -3.810614 5.619619 Prob. 0.0041 0.0003 260.8157 12.72778 12.80012 31.58011 0.000326 0.778216 Mean dependent var 331.3636 0.753574 S.D. dependent var 129.4724 Akaike info criterion 150867.9 Schwarz criterion -68.00278 F-statistic 1.142088 Prob(F-statistic)

Estimation Command:

===================== LS Y C X

Estimation Equation:

===================== Y = C(1) + C(2)*X

Substituted Coefficients: =====================

Y = -744.6350676 + 0.08825777732*X

由路径:Quick/Qstimate Equation,进入Equation Specification窗口,键入“y c x”,确认并“ok”,得样本回归估计结果,(注意:sample:21 31)

输出结果如下(表5.4):

Dependent Variable: Y Method: Least Squares Date: 04/14/08 Time: 10:27 Sample: 21 31

Included observations: 11

Variable C X

R-squared

Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat

Coefficient 666.3811 0.045779

Std. Error 911.2585 0.0278

t-Statistic 0.731276 1.0971

Prob. 0.4832 0.1352 3.4462 14.58557 14.65791 2.692786 0.135222

注意:期间是21 31 0.230295 Mean dependent var 2152.909 0.144772 S.D. dependent var 327.7867 Akaike info criterion 966997.0 Schwarz criterion -78.22063 F-statistic 2.743586 Prob(F-statistic)

Estimation Command:

===================== LS Y C X

Estimation Equation:

===================== Y = C(1) + C(2)*X

Substituted Coefficients: =====================

Y = 666.3810693 + 0.04577902024*X

(3)求F统计量值。基于表5.3和表5.4中残差平方和的数据,即Sum squared resid的值。由表5.3计算得到的残差平方和为方和为

21ie1i150867.9,由表5.4计算得到的残差平

2e966997.0,根据Goldfeld-Quanadt检验,F统计量为

e966997F6.41 (5.1)

150867.9e22i21i(4)判断。在0.05下,式(5.1)中分子、分母的自由度均为11,查F分布表得临界值为F(0.05)(112,112)3.18,因为F6.41F(0.05)(112,112)3.18,所以拒绝原假设,表明模型确实存在异方差。

(三)White检验

由表5.2估计结果,按路径view/residual tests/white heteroskedasticity(no cross terms or cross terms),进入White检验。根据White检验中辅助函数的构造,最后一项为变量的交叉乘积项,因为本例为一元函数,故无交叉乘积项,因此应选no cross terms,则辅助函数为

22xxvt (5.2) t01t2t

经估计出现White检验结果,见表5.5。

从表5.5可以看出,nR9.1026,由White检验知,在0.05下,查分布表,

22(2)5.9915(在(5.2)式中只有两项含有解释变量,故自由度为2)0得临界值.05,比较

222计算的统计量与临界值,因为nR9.10265.9915,所以拒绝原假设,不拒绝备择

假设,表明模型存在异方差。

表5.5

White Heteroskedasticity Test: F-statistic 5.819690

0.007699

Probability

Obs*R-squared Test Equation: Dependent Variable: RESID^2 Method: Least Squares Date: 04/14/08 Time: 11:02 Sample: 1 31 Included observations: 31 Variable C X X^2 R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat Coefficient 19975.98 -2.198632 0.000146 Std. Error 82774.93 8.094419 0.000176 t-Statistic 0.241329 -0.271623 0.830046 Prob. 0.8111 0.7879 0.4135 77875.67 25.17675 25.31553 5.819690 0.007699 9.102584 Probability 0.0105 0.293632 Mean dependent var 55881.73 0.243177 S.D. dependent var 67748.39 Akaike info criterion 1.29E+11 Schwarz criterion -387.2397 F-statistic 2.580140 Prob(F-statistic) 四、异方差性的修正 (一)加权最小二乘法(WLS)

w1t在运用WLS法估计过程中,我们分别选用了权数

111,w2i2,w3iXtXtXt。权

数的生成过程如下,由图5.4,的主菜单中点击“quick”下的“Estimation Equation”键,输入公式:Y C X,同时,点击“option”键,在菜单中“weight”后面的空白处输入“1/X”,

点击该键,输入权重

权重:可赋予不同形式,目标是消除异方差

下面仅给出用权数W1t的结果。

Dependent Variable: Y Method: Least Squares Date: 04/25/08 Time: 22:28 Sample: 1 31

Included observations: 31 Weighting series: 1/X Variable C X Weighted Statistics R-squared

Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat Unweighted Statistics R-squared

Adjusted R-squared S.E. of regression Durbin-Watson stat

Coefficient -742.4684 0.0751 Std. Error 71.91567 0.004347 t-Statistic -10.32415 20.696 Prob. 0.0000 0.0000 406.6195 13.40755 13.50006 426.2970 0.000000 846.7570 1741810.

0.786117 Mean dependent var 903.0766 0.778742 S.D. dependent var 191.2661 Akaike info criterion 10609. Schwarz criterion -205.8170 F-statistic 1.081175 Prob(F-statistic)

0.919023 Mean dependent var 1266.452 0.916231 S.D. dependent var 245.0763 Sum squared resid 1.074712

基于上述结果的E残差WHITE检验

White Heteroskedasticity Test: F-statistic Obs*R-squared

Test Equation:

Dependent Variable: STD_RESID^2 Method: Least Squares Date: 06/03/11 Time: 08:47 Sample: 1 31

Included observations: 31 Variable C X X^2

R-squared

Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat Coefficient 110552.8 -7.342468 0.000151

Std. Error 963.86 5.374822 0.000117

t-Statistic 2.011373 -1.366086 1.293401

Prob. 0.00 0.1828 0.20 44965.37 24.35786 24.49663 0.986241 0.385562 0.986241 Probability 2.040103 Probability

0.385562 0.360576

0.065810 Mean dependent var 34222.55 -0.000918 S.D. dependent var 44986.00 Akaike info criterion 5.67E+10 Schwarz criterion -374.68 F-statistic 1.733930 Prob(F-statistic)

TR2=31*0.065810=2.04011友情提示:范文可能无法思考和涵盖全面,供参考!最好找专业人士起草或审核后使用,感谢您的下载!

因篇幅问题不能全部显示,请点此查看更多更全内容

Copyright © 2019- yrrf.cn 版权所有 赣ICP备2024042794号-2

违法及侵权请联系:TEL:199 1889 7713 E-MAIL:2724546146@qq.com

本站由北京市万商天勤律师事务所王兴未律师提供法律服务