I'm trying to run etregress with an interaction. The interaction is between the treatment and a dummy variable.
In the manual (Treatment Reference, page 24, example 2 under etregress), it does an interaction with black but labels black as continuous. The passage from the manual is:
etregress wage age grade smsa i.union#c.(black tenure),
> treat(union = south black tenure) vce(robust)
If I label the interacting variable as an indicator or don't label it at all it will not estimate under MLE, and twostep indicates it is trying to estimate a parameter that it cannot estimate:
etregress wage age grade smsa i.union#black i.union#c.tenure, ///
> treat(union = south black tenure) twostep
Linear regression with endogenous treatment Number of obs = 1210
Estimator: two-step Wald chi2(8) = 520.96
Prob > chi2 = 0.0000
--------------------------------------------------------------------------------
| Coef. Std. Err. z P>|z| [95% Conf. Interval]
---------------+----------------------------------------------------------------
wage |
age | .1559505 .0194979 8.00 0.000 .1177352 .1941657
grade | .4205298 .0288255 14.59 0.000 .3640329 .4770268
smsa | .857708 .1292047 6.64 0.000 .6044714 1.110945
|
union#black |
0 1 | -.8234432 .1923978 -4.28 0.000 -1.200536 -.4463505
1 0 | 1.431043 .2834431 5.05 0.000 .8755048 1.986581
1 1 | 0 (omitted)
|
union#c.tenure |
0 | .1110202 .055226 2.01 0.044 .0027792 .2192612
1 | -.0252598 .082905 -0.30 0.761 -.1877506 .137231
|
union | 4.199704 1.053091 3.99 0.000 2.135683 6.263724
_cons | -4.839299 .5472206 -8.84 0.000 -5.911832 -3.766766
---------------+----------------------------------------------------------------
union |
south | -.4895032 .0933276 -5.24 0.000 -.6724221 -.3065844
black | .4397974 .0972261 4.52 0.000 .2492377 .6303572
tenure | .0997638 .0236575 4.22 0.000 .053396 .1461317
_cons | -.9679795 .0746464 -12.97 0.000 -1.114284 -.8216753
---------------+----------------------------------------------------------------
hazard |
lambda | -2.42826 .637324 -3.81 0.000 -3.677392 -1.179128
---------------+----------------------------------------------------------------
rho | -0.97401
sigma | 2.4930603
--------------------------------------------------------------------------------
I would normally worry about having Stata treat a dummy as continuous.
Is it then correct to label the interacting dummy as continuous?
If so, is there a natural interpretation of the coefficients? That is, how would I compare the effect of union on blacks versus whites?
Phil
In the manual (Treatment Reference, page 24, example 2 under etregress), it does an interaction with black but labels black as continuous. The passage from the manual is:
etregress wage age grade smsa i.union#c.(black tenure),
> treat(union = south black tenure) vce(robust)
If I label the interacting variable as an indicator or don't label it at all it will not estimate under MLE, and twostep indicates it is trying to estimate a parameter that it cannot estimate:
etregress wage age grade smsa i.union#black i.union#c.tenure, ///
> treat(union = south black tenure) twostep
Linear regression with endogenous treatment Number of obs = 1210
Estimator: two-step Wald chi2(8) = 520.96
Prob > chi2 = 0.0000
--------------------------------------------------------------------------------
| Coef. Std. Err. z P>|z| [95% Conf. Interval]
---------------+----------------------------------------------------------------
wage |
age | .1559505 .0194979 8.00 0.000 .1177352 .1941657
grade | .4205298 .0288255 14.59 0.000 .3640329 .4770268
smsa | .857708 .1292047 6.64 0.000 .6044714 1.110945
|
union#black |
0 1 | -.8234432 .1923978 -4.28 0.000 -1.200536 -.4463505
1 0 | 1.431043 .2834431 5.05 0.000 .8755048 1.986581
1 1 | 0 (omitted)
|
union#c.tenure |
0 | .1110202 .055226 2.01 0.044 .0027792 .2192612
1 | -.0252598 .082905 -0.30 0.761 -.1877506 .137231
|
union | 4.199704 1.053091 3.99 0.000 2.135683 6.263724
_cons | -4.839299 .5472206 -8.84 0.000 -5.911832 -3.766766
---------------+----------------------------------------------------------------
union |
south | -.4895032 .0933276 -5.24 0.000 -.6724221 -.3065844
black | .4397974 .0972261 4.52 0.000 .2492377 .6303572
tenure | .0997638 .0236575 4.22 0.000 .053396 .1461317
_cons | -.9679795 .0746464 -12.97 0.000 -1.114284 -.8216753
---------------+----------------------------------------------------------------
hazard |
lambda | -2.42826 .637324 -3.81 0.000 -3.677392 -1.179128
---------------+----------------------------------------------------------------
rho | -0.97401
sigma | 2.4930603
--------------------------------------------------------------------------------
I would normally worry about having Stata treat a dummy as continuous.
Is it then correct to label the interacting dummy as continuous?
If so, is there a natural interpretation of the coefficients? That is, how would I compare the effect of union on blacks versus whites?
Phil
