Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: RES: treatreg model with binary outcomes


From   "Henrique Neder" <hdneder@ufu.br>
To   <statalist@hsphsun2.harvard.edu>
Subject   st: RES: treatreg model with binary outcomes
Date   Tue, 20 Nov 2007 13:16:59 -0200

Kim

I do not know if I understand your question well. I think that I had a
similar problem: estimating the effect of a social policy (a government
program to combat poverty) on an indicator of food security. Participation
in the program is defined as a dummy variable (P) and Y is the indicator of
food security which is also a dummy. The first equation is a logit model: 

Ln (p / (1-p)) = beta0 + beta1 * P + beta2 * X1 + ... + betap * Xp + u (1) 

Where p is the probability of being in a situation of food security. 

P is a dummy of selection in the program 

Clearly here P is endogenous because participation in the program is
determined by the situation of food security of the household. Then I used a
second equation: 

Ln (q / (1-q)) = gamma0 + gamma1 * M1 + ... + gammap * Mp + v (2) 

Where q is the selection probability in the program and M1 ,..., Mp are
explanatory factors for this participation. This equation is the first stage
of an estimation by instrumental variables in two stages. I replaced the
estimate of q obtained through the logit model (2) in place of P variable in
the logit model (1).

I wonder if this helps you and also on the part of any list member if there
is any flaw in this econometric method.

Henrique Neder


-----Mensagem original-----
De: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] Em nome de
manturuk@email.unc.edu
Enviada em: segunda-feira, 19 de novembro de 2007 23:04
Para: statalist@hsphsun2.harvard.edu
Assunto: st: treatreg model with binary outcomes

Hi all!

I want to run a treatment effects model.  I have binary outcome 
variables for both parts of a two-stage model.  Essentially, I have 
this:

W=Y+x1...xi+e where Y=Z+x1...x1+e
My outcome of interest is W (binary) and it is predicted by Y (binary) 
and control variables x1...xi.  Y is predicted by an instrumental 
variable Z (continuous, unrelated to W) and the the same set of control 
variables x1...xi (predictive of both W and Y).

It is my understanding that the treatreg command is appropriate only 
for a continuous outcome variable.  I have tried looking at the 
biprobit command but I have not been able to figure out if that is the 
appropriate model to use.  If so, which model gets the x1---xi 
variables?
-- 
Kim Manturuk


**********************************************************

"Fluoridation is the most monstrously conceived and dangerous plot we 
have ever
had to face."

~Dr. Strangelove




*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


__________ Informação do NOD32 IMON 2666 (20071119) __________

Esta mensagem foi verificada pelo NOD32 sistema antivírus
http://www.eset.com.br



*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index