Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: AW: Heckman Selection Rule


From   mtavoni <mtavoni@Princeton.EDU>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: AW: Heckman Selection Rule
Date   Tue, 25 Aug 2009 10:34:17 -0400

Thanks Martin.

The background to my question is a meta-analysis of a number of economic numerical models simulating a number of climate change mitigation scenarios. Some models didn't report results for the more ambitious climate policies because they were exceeding a given maximum cost (others were just unable to find a solution). I thought of employing Heckman to correct for this selection bias but from what you say this might not be the right tool. The bias here is that I'm leaving out from the sample all the high cost observations. any suggestion on how to tackle it ?

thanks
m

Martin Weiss wrote:
<>


You may want to add a little background to your question (why do you want to
do this?), but my feeling is that
(1) the change in the inequality sign in the selection equation merely
changes the sign of the components of "g" (and the estimate of "rho")

(2) the constant that you want on the RHS will be absorbed into the constant
that -heckman- adds by default to the selection equation. Note that there is
a separate -noconstant- option for it in the -heckman- command...


Bottom line: The outcome equation - in which interest ultimately rests - is
unchanged...


***
vers 10.1

clear*
set obs 10000

//correlation
loc co 0.7

//draw errors
//correlation as in local `co'
matrix input correl = /* */ (1,`co' \ `co',1) drawnorm u1 u2, /* */ corr(correl) /* */ cstorage(full)

//covariates
gen x1=rnormal() gen x2=runiform()
gen x3=rchi2(4)

//outcome equation
gen y=2+3*x1+ 4*x3+u1

//selection equation
replace y= . if /* */ -1+2*x2+x1+u2<0 heckman y x1 x3 ,/* */ select(x1 x2)/* */ nolog two est store heck


//next round
drop y
//_new_ outcome equation
gen y=2+3*x1+4*x3+u1

//_new_ selection equation
//Constant on RHS and changed
//inequality sign

replace y= . if /* */ -1+2*x2+x1+u2>3 heckman y x1 x3 ,/* */ select(x1 x2)/* */ nolog two

est table heck .
***


HTH
Martin


-----Ursprüngliche Nachricht-----
Von: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] Im Auftrag von mtavoni
Gesendet: Dienstag, 25. August 2009 00:08
An: statalist@hsphsun2.harvard.edu
Betreff: st: Heckman Selection Rule

Dear all,

a quick question on sample selection estimation. In Heckman, STATA estimates the parameters in the model:

        (regression equation: y is depvar, x is varlist)
        y = xb + u_1

        (selection equation: Z is varlist_s)
        y observed if Zg + u_2 > 0


What if I want to change the rule for the selection equation to something
like:
  y observed if Zg + u_2 < Constant


thanks for helping out.

cheers
max

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

--
Massimo Tavoni

Research Associate, Princeton Environmental Institute
Guyot Hall, Princeton University, Princeton, NJ, USA

Senior Researcher, Fondazione Eni Enrico Mattei (FEEM)
Cso Magenta 63, 20123, Milano, Italy

office: 0016092588474
cell:   0016094963618
fax:    0016092581716



*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   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