1. 那个文件,请贴在do file editor下执行 【该文件算是有些许不完美,会有错误讯息,但不影响重要结果】
. webuse highschool,clear
. run "C:\DOCUME~1\ADMINI~1\LOCALS~1\Temp\STD03000000.tmp"
. ivqreg school weight (height = sex) ,q(0.1) l(90)
(0 observations deleted)
Initial Estimation: .1th Two Stage Quantile Regression Number of obs = 4071
------------------------------------------------------------------------------
school | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
height | -.0339053 7.84e-09 -4.3e+06 0.000 -.0339053 -.0339052
weight | .0066896 4.14e-09 1.6e+06 0.000 .0066896 .0066896
_cons | 15.04294 2.95e-06 5.1e+06 0.000 15.04293 15.04294
------------------------------------------------------------------------------
Grid search is in progress (200)
<istmt>: 3499 ivqr_initial() not found
r(3499);
2.为了探究上述结果,借用Wooldridge那本panel data书中的control funtion approach【CFA】
套用在分位回归上【这个国外有人用过】
. reg height sex weight
. predict rho2,r
. qreg school weight height rho2,q(0.1) l(90)
.1 Quantile regression Number of obs = 4071
Raw sum of deviations 3613.8 (about 1)
Min sum of deviations 3595.2 Pseudo R2 = 0.0051
------------------------------------------------------------------------------
school | Coef. Std. Err. t P>|t| [90% Conf. Interval]
-------------+----------------------------------------------------------------
weight | .0066896 2.58e-09 2.6e+06 0.000 .0066896 .0066896
height | -.0339053 4.99e-09 -6.8e+06 0.000 -.0339053 -.0339053
rho2 | .0339053 6.42e-09 5.3e+06 0.000 .0339053 .0339053
_cons | 15.04295 1.87e-06 8.1e+06 0.000 15.04295 15.04296
------------------------------------------------------------------------------
似乎两个的结果很像,很接近。
当然,如果被怀疑内生的解释变量为dummy二元,那么,指令ivqte就是首选罗!
我个人并没有实际用例子比较过" ivqte" 与 "运用CFA的分位回归" 。【按理会一样或类似】
希望我的演练对您能有所帮助,
祝 周日假期 顺心 开心 自在


雷达卡






京公网安备 11010802022788号







