[quote]peyzf 发表于 2009-8-16 11:53
而一般的reg模型可以直接用gfields命令。
如 xi:reg employer age i.sex i.status i.area
. xi:reg employer age i.sex i.status i.area
i.sex _Isex_1-2 (naturally coded; _Isex_1 omitted)
i.status _Istatus_1-3 (naturally coded; _Istatus_1 omitted)
i.area _Iarea_0-2 (naturally coded; _Iarea_0 omitted)
Source | SS df MS Number of obs = 1571
-------------+------------------------------ F( 5, 1565) = 14.25
Model | 269521194 5 53904238.8 Prob > F = 0.0000
Residual | 5.9198e+09 1565 3782605.38 R-squared = 0.0435
-------------+------------------------------ Adj R-squared = 0.0405
Total | 6.1893e+09 1570 3942228.42 Root MSE = 1944.9
------------------------------------------------------------------------------
employer | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
age | -19.11043 5.665632 -3.37 0.001 -30.22346 -7.997401
_Isex_2 | -424.8835 100.6651 -4.22 0.000 -622.3361 -227.4308
_Istatus_2 | -346.1651 135.7591 -2.55 0.011 -612.4539 -79.87627
_Istatus_3 | (dropped)
_Iarea_1 | -142.215 170.2102 -0.84 0.404 -476.079 191.649
_Iarea_2 | 487.5856 180.0431 2.71 0.007 134.4344 840.7368
_cons | 2761.409 272.2159 10.14 0.000 2227.463 3295.355
------------------------------------------------------------------------------
. gfields
Fields' decomposition of factor contributions
Factor | Share of SS
---------+---------------
age | 0.0075
_Isex_2 | 0.0098
_Istatus_2| 0.0046
_Istatus_3| 0.0000
_Iarea_1 | 0.0048
_Iarea_2 | 0.0169
residual | 0.9564
个人理解,以上系数反映的是各个变量对因变量影响的大小,这样一来,它与我们的标准化的系数回归方程参数有何差别?
另外,我想考查变量status对因变量的影响,能不能把Istatus_2前面的系数 0.0046与_Istatus_3的系数 0.0000简单相加?因为status为一离散变量,分为Istatus_2|、_Istatus_3两个虚拟变量。