《序列相关案例分析》由会员分享，可在线阅读，更多相关《序列相关案例分析（23页珍藏版）》请在金锄头文库上搜索。
1、第二十讲序列相关案例分析主要内容：p序列相关的侦查p序列相关的处理p考察我国1984年2004年进出口额与国内生产 总值之间的关系数据收集中国1989-2004年进进出口额额与国内生产总值产总值 单位：亿元 年份国内生产总值进出口额 198916,909.204,156.00 199018,547.905,560.10 199121,617.807,225.80 199226,638.109,119.60 199334,634.4011,271.00 199446,759.4020,381.90 199558,478.1023,499.90 199667,884.6024,133.80 199
2、774,462.6026,967.20 199878,345.2026,849.70 199982,067.5029,896.20 200089,468.1039,273.20 200197,314.8042,183.60 2002105,172.3051,378.20 2003117,390.2070,483.50 2004136,875.9095,539.10回归方程pDependent Variable: YpMethod: Least SquarespDate: 11/28/10 Time: 20:42pSample: 1989 2004pIncluded observations:
3、16ppVariable Coefficient Std. Errort-Statistic Prob. pp C -11935.44 4575.009-2.608834 0.0206p X 0.632955 0.06010810.53021 0.0000ppR-squared 0.887897 Mean dependent var 30494.93pAdjusted R-squared 0.879890 S.D. dependent var 25007.71pS.E. of regression 8666.900 Akaike info criterion 21.08888pSum squa
4、red resid 1.05E+09 Schwarz criterion 21.18545pLog likelihood -166.7110 Hannan-Quinn criter. 21.09382pF-statistic 110.8854 Durbin-Watson stat0.383500pProb(F-statistic) 0.000000p序列相关的侦查（1）图解法p（2）DW检验在给定 ， ，查表得，存在着正的序列相关pNull Hypothesis: E has a unit rootpExogenous: ConstantpLag Length: 3 (Automatic ba
5、sed on SIC, MAXLAG=3)p t-Statistic Prob.*ppAugmented Dickey-Fuller test statistic -1.358756 0.5657pTest critical values: 1% level -4.121990p 5% level -3.144920p 10% level -2.713751pp（3）BG检验Breusch-Godfrey Serial Correlation LM Test: F-statistic 23.01977 Prob. F(2,12)0.0001 Obs*R-squared12.69191 Prob
6、. Chi-Square(2) 0.0018Test Equation: Dependent Variable: RESID Method: Least Squares Date: 11/28/10 Time: 21:23 Sample: 1989 2004 Included observations: 16 Presample missing value lagged residuals set to zero.VariableCoefficient Std. Errort-StatisticProb. C-2373.817 2800.904-0.847518 0.4133 X0.05057
7、30.0440251.1487270.2731 RESID(-1)1.5764310.2778355.6739860.0001 RESID(-2)-0.572916 0.393145-1.457265 0.1707R-squared 0.793244 Mean dependent var 5.23E-12 Adjusted R-squared0.741556 S.D. dependent var 8373.021 S.E. of regression4256.629 Akaike info criterion 19.76266 Sum squared resid2.17E+08 Schwarz
8、 criterion19.95581 Log likelihood-154.1013 Hannan-Quinn criter.19.77255 F-statistic 15.34651 Durbin-Watson stat 1.997979 Prob(F-statistic)0.000208pBreusch-Godfrey Serial Correlation LM Test:pF-statistic 23.0197 Prob. F(2,12) 0.0001pObs*R-squared 12.69191 Prob. Chi-Square(2)0.0018p序列相关的处理 （1）可行性GLS法1
9、、基于d统计量的FGLS，生成 ，进行估计pDependent Variable: Y1pMethod: Least SquarespDate: 11/28/10 Time: 22:34pSample (adjusted): 1990 2004pIncluded observations: 15 after adjustmentspppVariable Coefficient Std. Error t-StatisticProb. p C -8403.595 2756.874 -3.0482340.0093p X1 0.977589 0.125228 7.8064760.0000ppR-squ
10、ared 0.824184 Mean dependent var 11108.13pAdjusted R-squared 0.810660 S.D. dependent var 10353.99pS.E. of regression 4505.354 Akaike info criterion 19.78749pSum squared resid 2.64E+08 Schwarz criterion19.88189pLog likelihood -146.4061 Hannan-Quinn criter.19.78648pF-statistic 60.94107 Durbin-Watson s
11、tat 0.725518pProb(F-statistic) 0.000003p（2）基于残差自回归的FGLS先估计 ，得到再进行GLSDependent Variable: E Method: Least Squares Date: 11/28/10 Time: 22:53 Sample (adjusted): 1990 2004 Included observations: 15 after adjustmentsVariableCoefficient Std. Error t-Statistic Prob. C1082.892 1442.118 0.750904 0.4661E11.03
12、8092 0.224787 4.618117 0.0005R-squared0.621289 Mean dependent var-359.2454 Adjusted R-squared 0.592158 S.D. dependent var8538.311 S.E. of regression5452.780 Akaike info criterion20.16920 Sum squared resid3.87E+08 Schwarz criterion 20.26361 Log likelihood-149.2690 Hannan-Quinn criter.20.16820 F-stati
13、stic 21.32701 Durbin-Watson stat0.626748 Prob(F-statistic)0.000482Dependent Variable: Y1 Method: Least Squares Date: 11/28/10 Time: 22:57 Sample (adjusted): 1990 2004 Included observations: 15 after adjustmentsVariableCoefficient Std. Error t-Statistic Prob. C-1389.632 2248.993 -0.617891 0.5473X11.1
14、53718 0.327927 3.518211 0.0038R-squared0.487741 Mean dependent var 5095.562 Adjusted R-squared 0.448337 S.D. dependent var 6718.627 S.E. of regression4990.195 Akaike info criterion 19.99190 Sum squared resid3.24E+08 Schwarz criterion 20.08631 Log likelihood-147.9393 Hannan-Quinn criter. 19.99090 F-statistic 12.37781 Durbin-Watson stat 0.689689 Prob(F-statistic)0.003780p3、奥科特-科克伦迭代法接上面的结果，再次计算Dependent Variable: E Method: Least Squares Date: 11/29/10 Time: 18:58 Sample (adjusted): 1990 2004 In
