I do have the same version as you.
One of the reasons I am using margeff rather than mfx is
that I want average marginal effects rather than marginal effects at the mean.
Below is a simple model, where jlpyrs is the number of
years in a jobless household (ranging from 0 to 7).
You will see that the marginal effects obtained with
margeff (for gender) are much smaller than the ones computed ‘manually’
(variables t0 to 7 below).
I did the same experiment with mlogit, probit and
xtprobit and I did get for these models the same values for both marginal
effects. That’s why I am wondering whether there is a bug in margeff after
oprobit.
. oprobit jlpyrs1 gender age
Iteration 0: log likelihood = -4482.9115
Iteration 1: log likelihood = -4426.1931
Iteration 2: log likelihood = -4426.0798
Iteration 3: log likelihood = -4426.0798
Ordered probit
regression
Number of obs = 5063
LR
chi2(2) = 113.66
Prob > chi2 = 0.0000
Log likelihood =
-4426.0798
Pseudo R2 = 0.0127
------------------------------------------------------------------------------
jlpyrs1
| Coef. Std.
Err. z
P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
gender | -.2000074
.0388654 -5.15 0.000
-.2761821 -.1238327
age
| .0184501 .001925
9.58 0.000 .0146772
.022223
-------------+----------------------------------------------------------------
/cut1
| 1.447378 .0794277 1.291703
1.603053
/cut2
| 1.698119
.0802337
1.540864 1.855374
/cut3
| 1.856428
.0810219
1.697628 2.015228
/cut4
| 1.988807
.0818218
1.828439 2.149174
/cut5
| 2.107942
.0826557
1.94594 2.269944
/cut6
| 2.220859
.0835591
2.057086 2.384632
/cut7
| 2.449037
.0858606
2.280753 2.61732
------------------------------------------------------------------------------
. margeff compute, dummies(gender) count
Average partial effects after oprobit
y = Pr(jlpyrs1)
------------------------------------------------------------------------------
variable
| Coef. Std.
Err. z
P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
0
|
gender |
.0101312 .0024466 4.14
0.000 .005336 .0149264
age
| -.0009372 .0001174 -7.99
0.000 -.0011672 -.0007072
-------------+----------------------------------------------------------------
1
|
gender |
-.0041591 .0009832 -4.23
0.000 -.0060862 -.002232
age
| .0003835 .0000484
7.92 0.000 .0002887
.0004784
-------------+----------------------------------------------------------------
2
|
gender |
-.0018261 .0004549 -4.01
0.000 -.0027178 -.0009344
age
| .0001687 .0000238
7.07 0.000 .000122
.0002154
-------------+----------------------------------------------------------------
3
|
gender |
-.0011467 .0002968 -3.86
0.000 -.0017284 -.000565
age
| .0001061 .0000163
6.51 0.000 .0000741
.000138
-------------+----------------------------------------------------------------
4
|
gender |
-.0007903 .0002117 -3.73
0.000 -.0012052 -.0003755
age
| .0000732 .0000121
6.05 0.000 .0000495
.0000969
-------------+----------------------------------------------------------------
5
|
gender |
-.0005767 .0001596 -3.61
0.000 -.0008894 -.0002639
age
| .0000535 9.46e-06
5.66 0.000
.000035 .000072
-------------+----------------------------------------------------------------
6
|
gender |
-.0007793 .0002127 -3.66
0.000 -.0011963 -.0003623
age
| .0000724 .0000121
5.99 0.000 .0000487
.0000961
-------------+----------------------------------------------------------------
7
|
gender |
-.0008529 .0002425 -3.52
0.000 -.0013281 -.0003777
age
| .0000797 .000014
5.70 0.000 .0000523
.0001072
------------------------------------------------------------------------------
. predict p0-p7
(option pr assumed; predicted probabilities)
. replace gender = 0
(2470 real changes made)
. predict a0-a7
(option pr assumed; predicted probabilities)
. replace gender = 1
(5063 real changes made)
. predict b0-b7
(option pr assumed; predicted probabilities)
. forval i = 0/7{
2. gen t`i' = b`i' -
a`i'
3. }
. sum t?
Variable
|
Obs Mean Std.
Dev. Min
Max
-------------+--------------------------------------------------------
t0 | 5063
.0559895 .0083395 .0390145 .0704771
t1 | 5063 -.0116048
.0004624 -.0120314 -.0101325
t2 | 5063 -.0072334
.0004286 -.0076197 -.0060018
t3 | 5063 -.0057244
.0005597 -.0063758 -.0043622
t4 | 5063
-.0047819 .000624 -.0056904 -.0033971
-------------+--------------------------------------------------------
t5 | 5063
-.0041453 .000661 -.0052198 -.0027662
t6 | 5063 -.0071023
.0014138 -.0096342 -.0043449
t7 | 5063
-.0153973 .004964 -.02606
-.0071643
