我们只做加权最小二乘法的检验
首先数据如下：
年份 | 文教科卫支出 | 国内生产总值 |
1989 | 553.33 | 16992.3 |
1990 | 617.29 | 18667.8 |
1991 | 708 | 21781.5 |
1992 | 792.96 | 26923.5 |
1993 | 957.77 | 35333.9 |
1994 | 1278.18 | 48197.9 |
1995 | 1467.06 | 60793.7 |
1996 | 1704.25 | 71176.6 |
1997 | 1903.59 | 78973 |
1998 | 2154.38 | 84402.3 |
1999 | 2408.06 | 89677.1 |
2000 | 2736.88 | 99214.6 |
2001 | 3361.02 | 109655.2 |
2002 | 3979.08 | 120332.7 |
2003 | 4505.51 | 135822.8 |
2004 | 5143.65 | 159878.3 |
2005 | 6104.18 | 183084.8 |
方法一：LS Y C X 回车 在得到的结果中选择Estimate然后Option 权重一栏填入 1/ABS（resid） 然后得到以残差的绝对值为权重的结果如下：
| Dependent Variable: Y |
| Method: Least Squares |
| Date: 05/04/14 Time: 23:28 |
| Sample: 1 17 |
| Included observations: 17 |
| Weighting series: 1/ABS(RESID) |
| White Heteroskedasticity-Consistent Standard Errors & Covariance |
| Variable | Coefficient | Std. Error | t-Statistic | Prob. |
| C | -309.8847 | 46.16158 | -6.713042 | 0.0000 |
| X | 0.033477 | 0.000421 | 79.52051 | 0.0000 |
| Weighted Statistics | | | | |
| R-squared | 0.999999 | Mean dependent var | 3256.772 |
| Adjusted R-squared | 0.999999 | S.D. dependent var | 12707.78 |
| S.E. of regression | 14.43846 | Akaike info criterion | 8.287798 |
| Sum squared resid | 3127.035 | Schwarz criterion | 8.385824 |
| Log likelihood | -68.44629 | F-statistic | 5270.260 |
| Durbin-Watson stat | 0.351367 | Prob(F-statistic) | 0.000000 |
| Unweighted Statistics | | | | |
| R-squared | 0.973707 | Mean dependent var | 2375.011 |
| Adjusted R-squared | 0.971954 | S.D. dependent var | 1699.910 |
| S.E. of regression | 284.6832 | Sum squared resid | 1215668. |
| Durbin-Watson stat | 0.262961 | | | |
方法二：GENR E=ABS(RESID)
GENR YE=Y/E
GENR XE=X/E
GENR CE=1/E
LS YE CE XE 回车得到结结果如下：
| Dependent Variable: YE |
| Method: Least Squares |
| Date: 05/04/14 Time: 23:32 |
| Sample: 1 17 |
| Included observations: 17 |
| Variable | Coefficient | Std. Error | t-Statistic | Prob. |
| CE | -309.8847 | 50.54989 | -6.130274 | 0.0000 |
| XE | 0.033477 | 0.000461 | 72.59656 | 0.0000 |
| R-squared | 0.999999 | Mean dependent var | 232.3115 |
| Adjusted R-squared | 0.999999 | S.D. dependent var | 906.4689 |
| S.E. of regression | 1.029921 | Akaike info criterion | 3.006973 |
| Sum squared resid | 15.91107 | Schwarz criterion | 3.104998 |
| Log likelihood | -23.55927 | Durbin-Watson stat | 0.351367
|
比较上述两个结果发现 大部分的结果都是一致的 但是为何Std. Error
S.E. of regression Sum squared resid 等值不一致 而且方法二也没有F检验
我想问问两种操作方法得到不同结论的原因 就大神指点！！
扫码加好友，拉您进群
全部版块
我的主页
收藏
