我使用的是面板数据,采用的是截面固定效应进行估计,结果如下:
Dependent Variable: LOG(U?) | | |
Method: Pooled EGLS (Cross-section weights) | |
Date: 07/19/11
Time: 19:29 | | |
Sample: 2000 2009 | | |
Included observations: 10 | | |
Cross-sections included: 30 | | |
Total pool (balanced) observations: 300 | |
Linear estimation after one-step weighting matrix |
White cross-section standard errors & covariance (d.f. corrected) |
| | | | |
| | | | |
Variable | Coefficient | Std. Error | t-Statistic | Prob. |
| | | | |
| | | | |
C | 0.691684 | 0.249469 | 2.772625 | 0.0060 |
LOG(TI?) | 0.064810 | 0.034821 | 1.861214 | 0.0638 |
LOG(SI?) | 0.213852 | 0.036747 | 5.819589 | 0.0000 |
LOG(MAR?) | 0.279092 | 0.021122 | 13.21326 | 0.0000 |
LOG(EDU?) | 0.051965 | 0.007196 | 7.221464 | 0.0000 |
LOG(TCI?) | 0.040666 | 0.009296 | 4.374695 | 0.0000 |
LOG(INF?) | 0.007372 | 0.011608 | 0.635129 | 0.5259 |
LOG(GDP?) | 0.035022 | 0.022746 | 1.539664 | 0.1248 |
| Effects Specification | | |
| | | | |
| | | | |
Cross-section fixed (dummy variables) | |
| | | | |
| | | | |
| Weighted Statistics | | |
| | | | |
| | | | |
R-squared | 0.993212 | Mean dependent var | 6.153570 |
Adjusted R-squared | 0.992282 | S.D. dependent var | 2.935491 |
S.E. of regression | 0.049947 | Sum squared resid | 0.656107 |
F-statistic | 1068.889 | Durbin-Watson stat | 0.689289 |
Prob(F-statistic) | 0.000000 | | | |
| | | | |
| | | | |
| Unweighted Statistics | | |
| | | | |
| | | | |
R-squared | 0.972199 | Mean dependent var | 3.726917 |
Sum squared resid | 0.786521 | Durbin-Watson stat | 0.544177 |
这是刚开始的估计结果,我发现D-W值只有0.54417,说明存在自相关,因此我加入一阶自回归ar(1),得到的结果如下:
Dependent Variable: LOG(U?) | | |
Method: Pooled EGLS (Cross-section weights) | |
Date: 07/19/11
Time: 19:30 | | |
Sample (adjusted): 2001 2009 | | |
Included observations: 9 after adjustments | |
Cross-sections included: 30 | | |
Total pool (balanced) observations: 270 | |
Iterate coefficients after one-step weighting matrix |
White cross-section standard errors & covariance (d.f. corrected) |
Convergence achieved after 20 total coef iterations |
| | | | |
| | | | |
Variable | Coefficient | Std. Error | t-Statistic | Prob. |
| | | | |
| | | | |
C | 1.981029 | 0.373473 | 5.304347 | 0.0000 |
LOG(TI?) | -0.031413 | 0.051841 | -0.605961 | 0.5451 |
LOG(SI?) | 0.019824 | 0.061881 | 0.320352 | 0.7490 |
LOG(MAR?) | 0.068068 | 0.038698 | 1.758972 | 0.0799 |
LOG(EDU?) | -0.002807 | 0.014675 | -0.191287 | 0.8485 |
LOG(TCI?) | -0.002943 | 0.007582 | -0.388197 | 0.6982 |
LOG(INF?) | -0.025982 | 0.019747 | -1.315753 | 0.1896 |
LOG(GDP?) | 0.169399 | 0.026599 | 6.368649 | 0.0000 |
AR(1) | 0.859268 | 0.023148 | 37.12131 | 0.0000 |
| Effects Specification | | |
| | | | |
| | | | |
Cross-section fixed (dummy variables) | |
| | | | |
| | | | |
| Weighted Statistics | | |
| | | | |
| | | | |
R-squared | 0.998136 | Mean dependent var | 7.493801 |
Adjusted R-squared | 0.997839 | S.D. dependent var | 3.619287 |
S.E. of regression | 0.029334 | Sum squared resid | 0.199637 |
F-statistic | 3357.803 | Durbin-Watson stat | 2.025999 |
Prob(F-statistic) | 0.000000 | | | |
| | | | |
| | | | |
虽然得到的结果不存在自相关了,但是自变量都变得不显著了,请问各位高手这中情况应该怎么办?
我的数据都是来自于统计年鉴,只有少量数据缺失,使用内插法填补了一下。而且按照经济学的知识,那几个变量正常来说都应该显著的,为什么加入一阶自相关项后就不显著呢?是不是因为存在内生性问题呢?希望大家给予指点,非常感谢!