第六章练习题及参考解答
6.1表6.5是中国1985-2016年货物进出口贸易总额(????)与国内生产总值(????)的数据。 表6.5 中国进出口贸易总额和国内生产总值 单位:亿元 年份 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 货物进出口贸易总额(Y) 2066.71 2580.4 3084.2 3821.8 4155.9 5560.12 7225.75 9119.62 11271.02 20381.9 23499.94 24133.86 26967.24 26849.68 29896.23 39273.25 国内生产总值(X) 9098.9 10376.2 12174.6 15180.4 17179.7 18872.9 22005.6 27194.5 35673.2 48637.5 61339.9 71813.6 79715.0 85195.5 90564.4 100280.1 年份 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 货物进出口贸易总额(Y) 42183.62 51378.15 70483.45 95539.09 116921.77 140974.74 166924.07 179921.47 150648.06 201722.34 236401.95 244160.21 258168.89 264241.77 245502.93 243386.46 国内生产总值(X) 110863.1 121717.4 137422.0 161840.2 187318.9 219438.5 270232.3 319515.5 349081.4 413030.3 489300.6 540367.4 595244.4 643974.0 689052.1 740598.7 资料来源:《中国统计年鉴2017》 (1)建立货物进出口贸易总额的对数ln????对国内生产总值的对数ln????的回归方程; (2)检测模型的自相关性;
(3)采用广义差分法处理模型中的自相关问题。
【练习题6.1参考解答】
回归结果
???=?2.714792+1.152178???????? ??????
(0.316996) (0.027331) ??=?8.5641 42.15675
??2=0.9834 F=1777.192 DW=0.3069
自相关检验 ①图示法
图1、2 ?????1与????的散点图以及模型残差图
由上面两个图可以发现模型残差存在惯性表现,很可能存在正自相关。 ②DW检验
由回归结果可知DW统计量为0.3069,同时n=32,k=1,在0.05的显著性水平下,????=1.37,????=1.50,因而模型中存在正相关。
③BG检验
阶数 AIC SIC 5 -1.275502 -0.954873 4 -1.287655 -1.012829 3 -1.276954 -1.047933 2 -1.338140 -1.154923 滞后阶数从5阶减小到2阶,AIC及SIC达到最小时,滞后阶数为2阶,此时n??2=
22.57582,已知??20.05(2)=5.99,n??2=22.56454>5.99,同时P值为0.0000,在0.05的显著性水平下拒绝原假设,即存在自相关。
表2 BG检验2阶回归结果
自相关补救
①DW反算法求ρ
由DW=0.3069,可知???=1?
????2
=1?
0.30692
=0.84655,可得广义差分方程:
?????????0.84655?????????1=??1(1?0.84655)+??2(?????????0.84655?????????1)+????
表3 广义差分结果-DW反算法
DW检验:由回归结果可知DW统计量为1.6284,同时n=31,k=1,在0.05的显著性水平下,????=1.36,????=1.50,即已消除自相关。 BG检验: 阶数 AIC SIC 5 -1.260821 -0.937017 4 -1.273263 -0.995718 3 -1.337004 -1.105716 2 -1.369938 -1.184908 滞后阶数从5阶减小到2阶,AIC及SIC达到最小时,滞后阶数为2阶,此时n??2=
0.945667,已知??20.05(2)=5.99,n??2=0.945667<5.99,同时P值为0.6232,在0.05的显著性水平下不拒绝原假设,即已消除自相关。
表4 广义差分BG检验2阶回归结果
?0.153154
则可知,??1=1?0.84655=?0.998071
???=?0.998071+1.009608???????? 最终模型为:ln??
②残差过原点回归求ρ
Dependent Variable: E Method: Least Squares Date: 02/07/18 Time: 20:48
Sample (adjusted): 1986 2016
Included observations: 31 after adjustments
Variable E(-1)
R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat
Coefficient 0.902706
Std. Error 0.108990
t-Statistic 8.282442
Prob. 0.0000 0.004999 0.206153 -1.477927 -1.431670 -1.462848
0.695552 Mean dependent var 0.695552 S.D. dependent var 0.113749 Akaike info criterion 0.388162 Schwarz criterion 23.90787 Hannan-Quinn criter. 1.579983
表5 残差序列过原点回归结果
回归结果为:????=0.902706?????1,可知???=0.902706。
进而得广义差分方程:
ln?????0.902706?????????1=??1(1?0.902706)+??2(?????????0.902706?????????1)+????
Dependent Variable: LNY-0.902706*LNY(-1) Method: Least Squares Date: 02/07/18 Time: 20:51 Sample (adjusted): 1986 2016
Included observations: 31 after adjustments
Variable C
LNX-0.902706*LNX(-1) R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood F-statistic Prob(F-statistic)
Coefficient -0.005775 0.939897
Std. Error 0.215666 0.170867
t-Statistic -0.026778 5.500756
Prob. 0.9788 0.0000 1.175339 0.158003 -1.470776 -1.378261 -1.440619 1.744560
0.510617 Mean dependent var 0.493742 S.D. dependent var 0.112422 Akaike info criterion 0.366521 Schwarz criterion 24.79703 Hannan-Quinn criter. 30.25831 Durbin-Watson stat 0.000006
表6 广义差分-残差序列过原点回归结果
DW检验:由回归结果可知DW统计量为1.744560,同时n=31,k=1,在0.05的显著性水
平下,????=1.36,????=1.50,因而模型已不存在自相关。 BG检验: 阶数 AIC SIC 5 -1.248335 -0.924532 4 -1.254886 -0.977340 3 -1.318563 -1.087275 2 -1.356845 -1.171814 滞后阶数从5阶减小到2阶,AIC及SIC达到最小时,滞后阶数为2阶,此时n??2=
0.464615,已知??20.05(2)=5.99,n??2=0.464615<5.99,同时P值为0.7927,在0.05的显
著性水平下不拒绝原假设,即已消除自相关。
Breusch-Godfrey Serial Correlation LM Test: F-statistic Obs*R-squared Test Equation:
Dependent Variable: RESID Method: Least Squares Date: 02/07/18 Time: 21:30 Sample: 1986 2016 Included observations: 31
Presample missing value lagged residuals set to zero.
Variable C
LNX-0.902706*LNX(-1)
RESID(-1) RESID(-2)
R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood F-statistic Prob(F-statistic)
?0.005775
0.205411 Prob. F(2,27)
0.8156 0.7927
0.464615 Prob. Chi-Square(2)
Coefficient 0.003693 -0.003000 0.121196 -0.039349
Std. Error 0.223479 0.177227 0.194472 0.201437
t-Statistic 0.016525 -0.016926 0.623205 -0.195342
Prob. 0.9869 0.9866 0.5384 0.8466 4.02E-16 0.110532 -1.356845 -1.171814 -1.296530 1.970873
0.014988 Mean dependent var -0.094458 S.D. dependent var 0.115635 Akaike info criterion 0.361028 Schwarz criterion 25.03110 Hannan-Quinn criter. 0.136941 Durbin-Watson stat 0.937095
广义差分BG检验2阶回归结果
则可知,??1=1?0.902706=?0.059356
???=?0.059356+0.939897???????? 最终模型为:ln??
③德宾两步法求???
构建模型 ln????=??1(1???)+??2???????????2???????????1+???????????1+????
Dependent Variable: LNY Method: Least Squares Date: 02/07/18 Time: 21:43 Sample (adjusted): 1986 2016
Included observations: 31 after adjustments
Variable C
Coefficient -0.068979
Std. Error 0.496455
t-Statistic -0.138943
Prob. 0.8905