这是用eviews6.0做的最小二乘法回归，GRP为地区生产总值，inv为固定资产投资，popu为常住人口，wage为职工平均工资。
| GRP=C(1)*INV+C(2)*POPU+C(3)*WAGE+C(4) | | | | |
| | | | | |
| | Coefficient | Std. Error | t-Statistic | Prob. |
| | | | | |
| C(1) | 0.717645 | 0.325238 | 2.206523 | 0.0695 |
| C(2) | 2.271141 | 1.978758 | 1.147761 | 0.2948 |
| C(3) | 0.142881 | 0.020348 | 7.021987 | 0.0004 |
| C(4) | -3507.746 | 2291.073 | -1.53105 | 0.1766 |
| | | | | |
| R-squared | 0.998674 | Mean dependent var | | 12198.08 |
| Adjusted R-squared | 0.998011 | S.D. dependent var | | 4647.608 |
| S.E. of regression | 207.2554 | Akaike info criterion | | 13.79496 |
| Sum squared resid | 257728.8 | Schwarz criterion | | 13.91599 |
| Log likelihood | -64.97478 | Hannan-Quinn criter. | | 13.66218 |
| F-statistic | 1506.58 | Durbin-Watson stat | | 1.798928 |
| Prob(F-statistic) | 0 | | | |
popu通不过显著性检验。我想是存在多重共线性吧，于是用了差分法，结果如下
| GRP=C(1)*INV1+C(2)*POPU1+C(3)*WAGE1+C(4) | | | | |
| | | | | |
| | Coefficient | Std. Error | t-Statistic | Prob. |
| | | | | |
| C(1) | 9.756399 | 3.981491 | 2.450439 | 0.0579 |
| C(2) | -23.29078 | 49.28234 | -0.472599 | 0.6564 |
| C(3) | 1.366622 | 0.439229 | 3.111409 | 0.0265 |
| C(4) | 43.64928 | 5882.853 | 0.00742 | 0.9944 |
| | | | | |
| R-squared | 0.700099 | Mean dependent var | | 12883.07 |
| Adjusted R-squared | 0.520158 | S.D. dependent var | | 4361.391 |
| S.E. of regression | 3021.162 | Akaike info criterion | | 19.16577 |
| Sum squared resid | 45637098 | Schwarz criterion | | 19.25343 |
| Log likelihood | -82.24598 | Hannan-Quinn criter. | | 18.97661 |
| F-statistic | 3.890721 | Durbin-Watson stat | | 1.577845 |
| Prob(F-statistic) | 0.088874 | | | |
看到p值也是醉了……
计量学渣在此求教各位大神！三个问题：1、差分法是不是也有无法解决多重共线性的时候？2、像现在这种情况下是不是只能把popu剔除呢，因为剔除后f值p值什么的都好美。3、但是理论上人口对地区生产总值的贡献应该是很大的，为什么会出现这种不符合理论的情况呢？
扫码加好友，拉您进群
全部版块
我的主页
收藏
