New package on SSC archive: rbiprobit – Stata module to estimate recursive bivariate probit regressions
Thanks to Kit Baum, the new package rbiprobit is now available on SSC and can be installed by typing
rbiprobit fits a recursive bivariate probit regression using maximum likelihood estimation. The model involves an outcome equation and a treatment equation, whereas the treatment variable is an endogenous explanatory variable of the outcome equation. Both dependent variables have to be binary and the model assumes that the error terms have a bivariate normal distribution.
Check out the GitHub Page of the new package for more background information, literature and examples.
If you find bugs, have problems running the command or request new features, you can DM on Twitter (@cobanomics) or open an issue on the GitHub repo. Updates to the package will be more frequent via GitHub and less frequent via SSC.
Syntax
After fitting the model, you can compute various treatment effects and marginal effects of the independent variables. Moreover, a peculiarity of the recursive bivariate probit model is also available: you can decompose the marginal effects of independent variables that are included in both equations.
Example for treatment effects
Computation of average treatment effect (ate), average treatment effect on treated (atet), and average treatment effect on conditional probability (atec)
Example for marginal effects
Computation of total, direct and indirect marginal effects on joint probability p11
Code:
ssc install rbiprobit
Check out the GitHub Page of the new package for more background information, literature and examples.
If you find bugs, have problems running the command or request new features, you can DM on Twitter (@cobanomics) or open an issue on the GitHub repo. Updates to the package will be more frequent via GitHub and less frequent via SSC.
Syntax
Code:
rbiprobit depvar [=] [indepvars] [if] [in] [weight], endogenous(depvar_en [=] [indepvars_en] [, enopts]) [options]
Example for treatment effects
Computation of average treatment effect (ate), average treatment effect on treated (atet), and average treatment effect on conditional probability (atec)
Code:
. webuse class10, clear
(Class of 2010 profile)
. rbiprobit graduate = income i.roommate i.hsgpagrp, ///
> endog(program = i.campus i.scholar income i.hsgpagrp)
. rbiprobit tmeffects, tmeffect(ate)
Treatment effect Number of obs = 2,500
Model VCE : OIM
Expression : Pr(graduate=1), predict(pmarg1)
Effect : Average treatment effect
dydx w.r.t. : 1.program
------------------------------------------------------------------------------
| Delta-method
| dy/dx Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
ate | .0981665 .0476222 2.06 0.039 .0048287 .1915044
------------------------------------------------------------------------------
. rbiprobit tmeffects, tmeffect(atet)
. rbiprobit tmeffects, tmeffect(atec)
Example for marginal effects
Computation of total, direct and indirect marginal effects on joint probability p11
Code:
. rbiprobit margdec, dydx(income) effect(total) predict(p11)
Average marginal effects Number of obs = 2,500
Model VCE : OIM
Expression : Pr(graduate=1,program=1), predict(p11)
dy/dx w.r.t. : income
------------------------------------------------------------------------------
| Delta-method
| dy/dx Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
income | .0032146 .002856 1.13 0.260 -.0023831 .0088123
------------------------------------------------------------------------------
. rbiprobit margdec, dydx(income) effect(direct) predict(p11)
. rbiprobit margdec, dydx(income) effect(indirect) predict(p11)
Comment