[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
Re: st: interpreting treatment effect model results
Are you using only the `treated' sample in your outcome regression? If so,
then a positive g means that those with higher outcome also are more
likely to be selected.
Be sure wheather your true intention is to estimate an
endonenous switching model, or only correct for sample selection bias.
On Thu, 17 Oct 2002, John Phillips wrote:
> I need help interpreting the results of estimating a treatment effects model
> in which
> Outcome equation: y = a + b1X1 + b2X2 + dT + gL + e
> where T is the treatment and L is selectivity correction variable that is
> produced via probit estimation of the following model:
> T = 1 if T* > 0
> T* = m + N1X1 + N2X2 + u.
> I have two questions.
> 1. I find that d is significantly negative and g is significantly positive.
> What is the economic interpretation of this result? My understanding is
> that d estimates the average treatment effect of T on y and that a t-test of
> g is interpreted as a test of exogeneity, which can be rejected when g is
> significant. Does a positive g tell me anything else?
> 2. X1 appears in both the outcome and selection equations. What is the
> interpretation of b1, the coefficient on X1, in the outcome equation? From
> Greene (5th edition, p.783), the interpretation in a more general sample
> selection setting is that b1 captures the marginal effect of a change in X1
> on y only in addition to the effect that the change in X1 has on the
> likelihood that T* > 0. Does that interpretation hold in my setting?
> John Phillips
> Department of Accounting
> School of Business
> University of Connecticut
> 2100 Hillside Road, Unit 1041
> Storrs, CT 06269-1041
> Telephone: 860/486-2789
> Fax: 860/486-4838
> e-mail: firstname.lastname@example.org
> * 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/
* For searches and help try: