Dear Statalist members,
I have a question regarding the interpretation of an output after the "churdle" command.
I know that most authors report marginal effects as outputs since the coefficients cannot be interpreted as unconditional effects. Since my regression includes an interaction term with a dummy I can discuss marginal effects for both cases, but reporting them in a table is not very handy. (Or am I wrong and there exist a good way to do this?)
Therefore, I would like to report the standard output and somehow interpret the respective coefficients. My understanding is that the estimates of the linear model are conditional estimates (conditional on selection). Is that correct?
Let us consider the example of the stata manual that models the decision to exercise or not as a function of commute, whours, and age. These variables are written in select(). Once a decision to exercise is made, the time an individual exercises is modeled as a linear function of age, smoke, distance, and single.
For instance, if the coefficient of age in the linear model is .0015116, can I say that given the individual is selected (given the individual exercises), a one unit increase of age, exercise hours increase .0015116 units? Is this interpretation correct?
Thank you very much!!!
I have a question regarding the interpretation of an output after the "churdle" command.
I know that most authors report marginal effects as outputs since the coefficients cannot be interpreted as unconditional effects. Since my regression includes an interaction term with a dummy I can discuss marginal effects for both cases, but reporting them in a table is not very handy. (Or am I wrong and there exist a good way to do this?)
Therefore, I would like to report the standard output and somehow interpret the respective coefficients. My understanding is that the estimates of the linear model are conditional estimates (conditional on selection). Is that correct?
Let us consider the example of the stata manual that models the decision to exercise or not as a function of commute, whours, and age. These variables are written in select(). Once a decision to exercise is made, the time an individual exercises is modeled as a linear function of age, smoke, distance, and single.
Code:
. use http://www.stata-press.com/data/r14/fitness
.
. churdle linear hours age i.smoke distance i.single, select(commute whours age) ll(0)
Iteration 0: log likelihood = -23657.236
Iteration 1: log likelihood = -23344.182
Iteration 2: log likelihood = -23340.051
Iteration 3: log likelihood = -23340.044
Iteration 4: log likelihood = -23340.044
Cragg hurdle regression Number of obs = 19,831
LR chi2(6) = 9059.26
Prob > chi2 = 0.0000
Log likelihood = -23340.044 Pseudo R2 = 0.1625
------------------------------------------------------------------------------
hours | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
hours |
age | .0015116 .000763 1.98 0.048 .0000162 .003007
|
smoke |
smoking | -1.06646 .0460578 -23.15 0.000 -1.156731 -.9761879
distance | -.1333868 .0126344 -10.56 0.000 -.1581497 -.1086238
|
single |
single | .9940893 .0258775 38.42 0.000 .9433703 1.044808
_cons | .9138855 .0396227 23.06 0.000 .8362264 .9915447
-------------+----------------------------------------------------------------
selection_ll |
commute | -.2953345 .0624665 -4.73 0.000 -.4177666 -.1729024
whours | .0022974 .0069306 0.33 0.740 -.0112864 .0158811
age | -.0485347 .0006501 -74.65 0.000 -.049809 -.0472604
_cons | 2.649945 .0499795 53.02 0.000 2.551987 2.747903
-------------+----------------------------------------------------------------
lnsigma |
_cons | .0083199 .0099648 0.83 0.404 -.0112107 .0278506
-------------+----------------------------------------------------------------
/sigma | 1.008355 .010048 .9888519 1.028242
------------------------------------------------------------------------------
Thank you very much!!!
Comment