It is simple to do this is in Stata.
First of all you have to notice a mathematical equivalence:
It is equivalent to write
P1=x1*beta1+e1 if delta=1,
P2=x2*beta2+e2 if delta=0,
And to write one "fully interacted model". A fully inteacted model is a
model with one equation with all the observations in which all regressors
are multiplied by a dummy variable (and the constant is omitted). Call the
dummies d1 and d2
P= alpha1*d1+ alpha2*d2+ d1*x*beta1 + d2*x*beta2 + e
Remember to specify the noconstant option
Now you can use the command
-heckman- in Stata
Ciao
Renzo
----------------------------------------------------------------------------
----
*From <fukugawa-nobuya@rieti.go.jp>
To statalist@hsphsun2.harvard.edu
Subject Re: st: switching reg without using switchr
Date Mon, 1 Dec 2003 15:00:30 +0900 (JST)
Dear all,
I have dataset on firms' performance for both participants in public scheme
and non-participants and need to estimate treatment effects controlling for
endogeneous sample selection.
A main equation is specified as follows,
P1=x1*beta1+e1 if delta=1, where P1 is performance of participants
P2=x2*beta2+e2 if delta=0, where P2 is performance of non-participants
A regime equation is specified as follows,
S=Z*gamma+u, where S is a latent variable.
delta=1 if S>0, =zero otherwise, where delta is an observable dummy.
A switchr program downloadable at ssc does not seem to help.
It conducts ols instead of probit for a regime equation and yeilds the
results with the same number of observations for both of main equations for
P1 and P2.
Does anybody know how to conduct switching regression without using switchr?
Thank you in advance.
*
* 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/