谢谢蓝色版主。但是还有问题不明白。我用xtprobit 帮助文件里的数据和命令做了一下,结果如下。
先做了xtprobit ....,pa 然后用mfx compute计算出的dy/dx 与之前回归的系数不同,这种结果应该是我想要的。可是我想做的是随机效应模型。接着用您给的命令margins,dydx(*),可是提示命令不正确。
然后我做了xtprobit....,re,用mfx compute计算出的dy/dx 与之前回归的系数相同,这与我用自己的数据做出的结果是一样的。
. webuse union
(NLS Women 14-24 in 1968)
. xtprobit union age grade not_smsa south southXt, pa
Iteration 1: tolerance = .04796083
Iteration 2: tolerance = .00352657
Iteration 3: tolerance = .00017886
Iteration 4: tolerance = 8.654e-06
Iteration 5: tolerance = 4.150e-07
GEE population-averaged model Number of obs = 26200
Group variable: idcode Number of groups = 4434
Link: probit Obs per group: min = 1
Family: binomial avg = 5.9
Correlation: exchangeable max = 12
Wald chi2(5) = 241.66
Scale parameter: 1 Prob > chi2 = 0.0000
------------------------------------------------------------------------------
union | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
age | .0031597 .0014678 2.15 0.031 .0002829 .0060366
grade | .0329992 .0062334 5.29 0.000 .020782 .0452163
not_smsa | -.0721799 .0275189 -2.62 0.009 -.1261159 -.0182439
south | -.409029 .0372213 -10.99 0.000 -.4819815 -.3360765
southXt | .0081828 .002545 3.22 0.001 .0031946 .0131709
_cons | -1.184799 .0890117 -13.31 0.000 -1.359259 -1.01034
------------------------------------------------------------------------------
. mfx compute
Marginal effects after xtprobit
y = normprob(xb) (predict)
= .2048386
------------------------------------------------------------------------------
variable | dy/dx Std. Err. z P>|z| [ 95% C.I. ] X
---------+--------------------------------------------------------------------
age | .0008973 .00042 2.15 0.032 .00008 .001715 30.4322
grade | .0093715 .00177 5.30 0.000 .005904 .012839 12.7615
not_smsa*| -.0202326 .00761 -2.66 0.008 -.035144 -.005321 .283702
south*| -.1125259 .00983 -11.45 0.000 -.131795 -.093257 .413015
southXt | .0023238 .00072 3.22 0.001 .000908 .003739 3.96874
------------------------------------------------------------------------------
(*) dy/dx is for discrete change of dummy variable from 0 to 1
. margins, dydx(*)
unrecognized command: margins
r(199);
. xtprobit union age grade not_smsa south southXt,re
Fitting comparison model:
Iteration 0: log likelihood = -13864.23
Iteration 1: log likelihood = -13548.436
Iteration 2: log likelihood = -13547.308
Iteration 3: log likelihood = -13547.308
Fitting full model:
rho = 0.0 log likelihood = -13547.308
rho = 0.1 log likelihood = -12239.207
rho = 0.2 log likelihood = -11591.449
rho = 0.3 log likelihood = -11212.156
rho = 0.4 log likelihood = -10982.153
rho = 0.5 log likelihood = -10853.488
rho = 0.6 log likelihood = -10809.372
rho = 0.7 log likelihood = -10866.13
Iteration 0: log likelihood = -10808.301
Iteration 1: log likelihood = -10599.612
Iteration 2: log likelihood = -10552.389
Iteration 3: log likelihood = -10552.327
Iteration 4: log likelihood = -10552.327
Random-effects probit regression Number of obs = 26200
Group variable: idcode Number of groups = 4434
Random effects u_i ~ Gaussian Obs per group: min = 1
avg = 5.9
max = 12
Wald chi2(5) = 220.73
Log likelihood = -10552.327 Prob > chi2 = 0.0000
------------------------------------------------------------------------------
union | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
age | .0046532 .0025134 1.85 0.064 -.000273 .0095795
grade | .0481056 .0099402 4.84 0.000 .0286232 .0675881
not_smsa | -.1401136 .0460446 -3.04 0.002 -.2303594 -.0498679
south | -.6428989 .0619944 -10.37 0.000 -.7644058 -.521392
southXt | .0130454 .0043914 2.97 0.003 .0044384 .0216525
_cons | -1.872166 .1455716 -12.86 0.000 -2.157481 -1.586851
-------------+----------------------------------------------------------------
/lnsig2u | .6105518 .0458732 .5206419 .7004617
-------------+----------------------------------------------------------------
sigma_u | 1.356999 .031125 1.297346 1.419395
rho | .6480667 .0104626 .6272979 .6682901
------------------------------------------------------------------------------
Likelihood-ratio test of rho=0: chibar2(01) = 5989.96 Prob >= chibar2 = 0.000
. browse
. mfx compute
Marginal effects after xtprobit
y = Linear prediction (predict)
= -1.3701643
------------------------------------------------------------------------------
variable | dy/dx Std. Err. z P>|z| [ 95% C.I. ] X
---------+--------------------------------------------------------------------
age | .0046532 .00251 1.85 0.064 -.000273 .009579 30.4322
grade | .0481056 .00994 4.84 0.000 .028623 .067588 12.7615
not_smsa*| -.1401136 .04604 -3.04 0.002 -.230359 -.049868 .283702
south*| -.6428989 .06199 -10.37 0.000 -.764406 -.521392 .413015
southXt | .0130454 .00439 2.97 0.003 .004438 .021652 3.96874
------------------------------------------------------------------------------
(*) dy/dx is for discrete change of dummy variable from 0 to 1
|