Title stata.

com
tobit postestimation — Postestimation tools for tobit

Description Syntax for predict Menu for predict Options for predict
Remarks and examples References Also see

Description
The following postestimation commands are available after tobit:

Command Description
contrast contrasts and ANOVA-style joint tests of estimates
estat ic Akaike’s and Schwarz’s Bayesian information criteria (AIC and BIC)
estat summarize summary statistics for the estimation sample
estat vce variance–covariance matrix of the estimators (VCE)
estat (svy) postestimation statistics for survey data
estimates cataloging estimation results
forecast1 dynamic forecasts and simulations
hausman Hausman’s specification test
lincom point estimates, standard errors, testing, and inference for linear
combinations of coefficients
linktest link test for model specification
lrtest2 likelihood-ratio test
margins marginal means, predictive margins, marginal effects, and average marginal
effects
marginsplot graph the results from margins (profile plots, interaction plots, etc.)
nlcom point estimates, standard errors, testing, and inference for nonlinear
combinations of coefficients
predict predictions, residuals, influence statistics, and other diagnostic measures
predictnl point estimates, standard errors, testing, and inference for generalized predictions
pwcompare pairwise comparisons of estimates
suest seemingly unrelated estimation
test Wald tests of simple and composite linear hypotheses
testnl Wald tests of nonlinear hypotheses
1
forecast is not appropriate with svy estimation results.
2
lrtest is not appropriate with svy estimation results.

1
 2 tobit postestimation — Postestimation tools for tobit

Syntax for predict

 
     
predict type
newvar if in , statistic nooffset
      
predict type stub* | newvarreg newvarsigma if in , scores

statistic Description
Main
xb linear prediction; the default
stdp standard error of the linear prediction
stdf standard error of the forecast
pr(a,b) Pr(a < yj < b)
e(a,b) E(yj | a < yj < b)
ystar(a,b) E(yj∗ ), yj∗ = max{a, min(yj , b)}

These statistics are available both in and out of sample; type predict . . . if e(sample) . . . if wanted only for
the estimation sample.
stdf is not allowed with svy estimation results.
where a and b may be numbers or variables; a missing (a ≥ .) means −∞, and b missing (b ≥ .)
means +∞; see [U] 12.2.1 Missing values.

Menu for predict

Statistics > Postestimation > Predictions, residuals, etc.

Options for predict


 Main
xb, the default, calculates the linear prediction.
stdp calculates the standard error of the prediction, which can be thought of as the standard error of
the predicted expected value or mean for the observation’s covariate pattern. The standard error
of the prediction is also referred to as the standard error of the fitted value.
stdf calculates the standard error of the forecast, which is the standard error of the point prediction
for 1 observation. It is commonly referred to as the standard error of the future or forecast value.
By construction, the standard errors produced by stdf are always larger than those produced by
stdp; see Methods and formulas in [R] regress postestimation.
pr(a,b) calculates Pr(a < xj b + uj < b), the probability that yj |xj would be observed in the
interval (a, b).
a and b may be specified as numbers or variable names; lb and ub are variable names;
pr(20,30) calculates Pr(20 < xj b + uj < 30);
pr(lb,ub) calculates Pr(lb < xj b + uj < ub); and
pr(20,ub) calculates Pr(20 < xj b + uj < ub).
a missing (a ≥ .) means −∞; pr(.,30) calculates Pr(−∞ < xj b + uj < 30);
pr(lb,30) calculates Pr(−∞ < xj b + uj < 30) in observations for which lb ≥ .
and calculates Pr(lb < xj b + uj < 30) elsewhere.
 tobit postestimation — Postestimation tools for tobit 3

b missing (b ≥ .) means +∞; pr(20,.) calculates Pr(+∞ > xj b + uj > 20);

pr(20,ub) calculates Pr(+∞ > xj b + uj > 20) in observations for which ub ≥ .
and calculates Pr(20 < xj b + uj < ub) elsewhere.
e(a,b) calculates E(xj b + uj | a < xj b + uj < b), the expected value of yj |xj conditional on
yj |xj being in the interval (a, b), meaning that yj |xj is truncated.
a and b are specified as they are for pr().
ystar(a,b) calculates E(yj∗ ), where yj∗ = a if xj b + uj ≤ a, yj∗ = b if xj b + uj ≥ b, and
yj∗ = xj b + uj otherwise, meaning that yj∗ is censored. a and b are specified as they are for pr().
nooffset is relevant only if you specified offset(varname). It modifies the calculations made by
predict so that they ignore the offset variable; the linear prediction is treated as xj b rather than
as xj b + offsetj .
scores calculates equation-level score variables.
The first new variable will contain ∂ ln L/∂(xj β).
The second new variable will contain ∂ ln L/∂σ .

Remarks and examples stata.com

Following Cong (2000), write the tobit model as

yi , if a < yi < b
(
yi∗ = a, if yi ≤ a
b, if yi ≥ b

yi is a latent variable; instead, we observe yi∗ , which is bounded between a and b if yi is outside
those bounds.
There are four types of marginal effects that may be of interest in the tobit model, depending on
the application:
1. The β coefficients themselves measure how the unobserved variable yi changes with respect
to changes in the regressors.
2. The marginal effects of the truncated expected value E(yi∗ |a < yi∗ < b) measure the changes
in yi with respect to changes in the regressors among the subpopulation for which yi is not
at a boundary.
3. The marginal effects of the censored expected value E(yi∗ ) describe how the observed
variable yi∗ changes with respect to the regressors.
4. The marginal effects of Pr(a < yi∗ < b) describe how the probability of being uncensored
changes with respect to the regressors.
In the next example, we show how to obtain each of these.
 4 tobit postestimation — Postestimation tools for tobit

Example 1
In example 3 of [R] tobit, we fit a two-limit tobit model of mpg on wgt.
. use http://www.stata-press.com/data/r13/auto
(1978 Automobile Data)
. generate wgt = weight/1000
. tobit mpg wgt, ll(17) ul(24)
Tobit regression Number of obs = 74
LR chi2(1) = 77.60
Prob > chi2 = 0.0000
Log likelihood = -104.25976 Pseudo R2 = 0.2712

mpg Coef. Std. Err. t P>|t| [95% Conf. Interval]

wgt -5.764448 .7245417 -7.96 0.000 -7.208457 -4.320438

_cons 38.07469 2.255917 16.88 0.000 33.57865 42.57072

/sigma 2.886337 .3952143 2.098676 3.673998

Obs. summary: 18 left-censored observations at mpg<=17

33 uncensored observations
23 right-censored observations at mpg>=24

tobit reports the β coefficients for the latent regression model. The marginal effect of xk on y is
simply the corresponding βk , because E(y|x) is linear in x. Thus a 1,000-pound increase in a car’s
weight (which is a 1-unit increase in wgt) would lower fuel economy by 5.8 mpg.
To estimate the means of the marginal effects on the expected value of the censored outcome,
conditional on weight being each of three values (2,000; 3,000; and 4,000 pounds), we type
. margins, dydx(wgt) predict(ystar(17,24)) at(wgt=(2 3 4))
Conditional marginal effects Number of obs = 74
Model VCE : OIM
Expression : E(mpg*|17<mpg<24), predict(ystar(17,24))
dy/dx w.r.t. : wgt
1._at : wgt = 2
2._at : wgt = 3
3._at : wgt = 4

Delta-method
dy/dx Std. Err. z P>|z| [95% Conf. Interval]

wgt
_at
1 -1.0861 .311273 -3.49 0.000 -1.696184 -.4760162
2 -4.45315 .4772541 -9.33 0.000 -5.388551 -3.51775
3 -1.412822 .3289702 -4.29 0.000 -2.057591 -.768052

The E(y ∗ |x) is nonlinear in x, so the marginal effect for a continuous covariate is not the same
as the change in y ∗ induced by a one-unit change in x. Recall that the marginal effect at a point
is the slope of the tangent line at that point. In our example, we estimate the mean of the marginal
effects for different values of wgt. The estimated mean of the marginal effects is −1.1 mpg for a
2,000 pound car; −4.5 mpg for a 3,000 pound car; and −1.4 mpg for a 4,000 pound car.
 tobit postestimation — Postestimation tools for tobit 5

To estimate the means of the marginal effects on the expected value of the truncated outcome at
the same levels of wgt, we type
. margins, dydx(wgt) predict(e(17,24)) at(wgt=(2 3 4))
Conditional marginal effects Number of obs = 74
Model VCE : OIM
Expression : E(mpg|17<mpg<24), predict(e(17,24))
dy/dx w.r.t. : wgt
1._at : wgt = 2
2._at : wgt = 3
3._at : wgt = 4

Delta-method
dy/dx Std. Err. z P>|z| [95% Conf. Interval]

wgt
_at
1 -1.166572 .0827549 -14.10 0.000 -1.328768 -1.004375
2 -2.308842 .4273727 -5.40 0.000 -3.146477 -1.471207
3 -1.288896 .0889259 -14.49 0.000 -1.463188 -1.114604

The mean of the marginal effects of a change in wgt on yi (which is bounded between 17 and 24)
is about −1.2 mpg for a 2,000 pound car; −2.3 mpg for a 3,000 pound car; and −1.3 for a 4,000
pound car.

References
Cong, R. 2000. sg144: Marginal effects of the tobit model. Stata Technical Bulletin 56: 27–34. Reprinted in Stata
Technical Bulletin Reprints, vol. 10, pp. 189–197. College Station, TX: Stata Press.
McDonald, J. F., and R. A. Moffitt. 1980. The use of tobit analysis. Review of Economics and Statistics 62: 318–321.

Also see
[R] tobit — Tobit regression
[U] 20 Estimation and postestimation commands