Dear stateliest-ers,
I am trying to replicate margins estimation after xttobit, based on the gsem routine. Here is what I have done.
The margins are estimated as follows:
I finally replicate the xttobit results with gsem:
Now, how can I get the same marginal effects I got after the xttobit?
Many thanks!
Rohit
I am trying to replicate margins estimation after xttobit, based on the gsem routine. Here is what I have done.
Code:
. webuse nlswork3
(National Longitudinal Survey. Young Women 14-26 years of age in 1968)
. xtset idcode year
panel variable: idcode (unbalanced)
time variable: year, 68 to 88, but with gaps
delta: 1 unit
. xttobit ln_wage union age grade not_smsa, ul(1.9)
Obtaining starting values for full model:
Iteration 0: log likelihood = -5570.3905
Iteration 1: log likelihood = -5434.9487
Iteration 2: log likelihood = -5433.7625
Iteration 3: log likelihood = -5433.7619
Fitting full model:
Iteration 0: log likelihood = -6935.1692
Iteration 1: log likelihood = -6877.0531
Iteration 2: log likelihood = -6875.2535
Iteration 3: log likelihood = -6875.252
Iteration 4: log likelihood = -6875.252
Random-effects tobit regression Number of obs = 19,224
Group variable: idcode Number of groups = 4,148
Random effects u_i ~ Gaussian Obs per group:
min = 1
avg = 4.6
max = 12
Integration method: mvaghermite Integration pts. = 12
Wald chi2(4) = 2770.47
Log likelihood = -6875.252 Prob > chi2 = 0.0000
------------------------------------------------------------------------------
ln_wage | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
union | .1482463 .0069843 21.23 0.000 .1345573 .1619353
age | .0104275 .0004148 25.14 0.000 .0096146 .0112404
grade | .0807764 .0022986 35.14 0.000 .0762713 .0852815
not_smsa | -.1454109 .0092189 -15.77 0.000 -.1634796 -.1273421
_cons | .3629613 .0319946 11.34 0.000 .3002531 .4256696
-------------+----------------------------------------------------------------
/sigma_u | .3094639 .0048855 63.34 0.000 .2998886 .3190393
/sigma_e | .2492985 .0018282 136.36 0.000 .2457152 .2528817
-------------+----------------------------------------------------------------
rho | .6064422 .0083153 .5900559 .6226421
------------------------------------------------------------------------------
0 left-censored observations
12,334 uncensored observations
6,890 right-censored observations
Code:
. margins, dydx(*) atmeans predict(ystar(.,1.9))
Conditional marginal effects Number of obs = 19,224
Model VCE : OIM
Expression : E(ln_wage*|ln_wage<1.9), predict(ystar(.,1.9))
dy/dx w.r.t. : union age grade not_smsa
at : union = .2342905 (mean)
age = 31.36366 (mean)
grade = 12.7642 (mean)
not_smsa = .2829796 (mean)
------------------------------------------------------------------------------
| Delta-method
| dy/dx Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
union | .1007414 .0047869 21.05 0.000 .0913592 .1101236
age | .0070861 .0002829 25.05 0.000 .0065317 .0076404
grade | .0548919 .0015846 34.64 0.000 .0517862 .0579977
not_smsa | -.0988145 .0063094 -15.66 0.000 -.1111807 -.0864483
------------------------------------------------------------------------------
Code:
. gsem ln_wage <- union age grade not_smsa RE[idcode], family(gaussian, rcensored(1.9)
> )
Fitting fixed-effects model:
Iteration 0: log likelihood = -11846.483
Iteration 1: log likelihood = -10432.472
Iteration 2: log likelihood = -10312.817
Iteration 3: log likelihood = -10312.543
Iteration 4: log likelihood = -10312.543
Refining starting values:
Grid node 0: log likelihood = -10058.171
Fitting full model:
Iteration 0: log likelihood = -10058.171 (not concave)
Iteration 1: log likelihood = -8744.4558 (not concave)
Iteration 2: log likelihood = -7819.9317 (not concave)
Iteration 3: log likelihood = -7098.5025
Iteration 4: log likelihood = -6918.365
Iteration 5: log likelihood = -6875.2693
Iteration 6: log likelihood = -6875.102
Iteration 7: log likelihood = -6875.102
Generalized structural equation model Number of obs = 19,224
Response : ln_wage Uncensored = 12,334
Upper limit : 1.9 Left-censored = 0
Family : Gaussian Right-censored = 6,890
Link : identity
Log likelihood = -6875.102
( 1) [ln_wage]RE[idcode] = 1
--------------------------------------------------------------------------------
| Coef. Std. Err. z P>|z| [95% Conf. Interval]
---------------+----------------------------------------------------------------
ln_wage <- |
union | .1482403 .006981 21.23 0.000 .1345578 .1619229
age | .0104287 .0004147 25.15 0.000 .0096159 .0112415
grade | .0808181 .0022927 35.25 0.000 .0763245 .0853117
not_smsa | -.1454202 .0092168 -15.78 0.000 -.1634848 -.1273557
|
RE[idcode] | 1 (constrained)
|
_cons | .3624916 .0319395 11.35 0.000 .2998912 .4250919
---------------+----------------------------------------------------------------
var(RE[idcode])| .0958442 .0030309 .0900841 .1019727
---------------+----------------------------------------------------------------
var(e.ln_wage)| .0621425 .0009115 .0603815 .0639549
--------------------------------------------------------------------------------
Many thanks!
Rohit