<>
The manual [R], p. 668, does state that it is possible to have the same set
of variables in both equations, but that in this case "...the model is
identified only by functional form, and the coefficients have no structural
interpretation."
Also note the very similar problem in [R], p. 649, for the -heckman-
command: "(Theoretically, we do not need such identifying variables, but
without them, we depend on functional form to identify the model. It would
be difficult for anyone to take such results seriously because the
functional-form assumptions have no firm basis in theory.)"
HTH
Martin
-----Ursprüngliche Nachricht-----
Von: [email protected]
[mailto:[email protected]] Im Auftrag von Celia Vera
Gesendet: Donnerstag, 26. November 2009 00:26
An: [email protected]
Betreff: st: heckprob
Hi statalisters
I am estimating a model with sample selection and the helpfile for heckprob
says that the selection equation should contain at least one variable that
is not in the outcome equation.
However, when I include the same variables in both the outcome and selection
equation, i get significant results.
Is my model not consistent by including the same variables?
Your help is appreciated.
Celia
----- Original Message ----
> From: Austin Nichols <[email protected]>
> To: [email protected]
> Sent: Wed, November 25, 2009 3:13:20 PM
> Subject: Re: st: RE: Retrieving statistics after pstest to use with estout
>
> Probably you just want one omnibus measure of balance (an F stat) but try
out:
>
> u http://pped.org/lalonde.dta, clear
> g age2=age^2
> g ed2=educ^2
> qui psmatch2 t age age2 educ ed2 re74
> pstest age age2 educ ed2 re74
> * test joint equality of all means
> preserve
> keep if _w<.
> expand _w
> qui reg t age age2 educ ed2 re74
> di e(F)
> qui hotelling age age2 educ ed2 re74, by(t)
> di (r(N)-r(k)-1)/(r(N)-2)/r(k)*r(T2)
> * or test vars one by one
> foreach v in age age2 educ ed2 re74 {
> qui ttest `v', by(t)
> di "`v' p-value:" r(p)
> }
> restore
> * or test one by one with paired data
> sort _id
> foreach v in age age2 educ ed2 re74 {
> qui g C`v'=`v'[_n1]
> qui ttest `v'=C`v'
> di "`v' p-value:" r(p)
> }
>
> On Wed, Nov 25, 2009 at 2:50 PM, Martin Weiss wrote:
> >
> > <>
> >
> > It is easy to -display- the call to the estimation command that -eststo-
> > picks up and -esttab- displays. Taking the example at
> > http://cameron.econ.ucdavis.edu/mmabook/mmaprograms.html:
> >
> >
> > *******
> > u http://cameron.econ.ucdavis.edu/mmabook/nswpsid.dta, clear
> >
> > psmatch2 TREAT AGE AGESQ EDUC EDUCSQ
> >
> > eststo psttest: pstest AGE AGESQ EDUC EDUCSQ, /*
> > */ sum t(TREAT)
> > esttab psttest
> >
> > di in r "`e(cmdline)'"
> > *******
> >
> > The only difficulty here is the temporary variable that they condition
on...
> >
> >
> > The return list from -pstest- contains only -scalar-s, so you would have
to
> > -trace- the command to see how you can replicate the calculation of your
> > desired results...
> >
> >
> > HTH
> > Martin
> >
> >
> > -----Original Message-----
> > From: [email protected]
> > [mailto:[email protected]] On Behalf Of Peer Skov
> > Sent: Mittwoch, 25. November 2009 19:12
> > To: [email protected]
> > Subject: st: Retrieving statistics after pstest to use with estout
> >
> > Dear listers,
> > After running a psmatch2 estimation I use pstest to assess the
balancing. I
> > would like to set-up an .rtf table showing these t-statistics and the
> > achieved percentage bias reduction for each variable. To avoid
copy-typing I
> > usually use estout/esttab, however, I haven't been able to reproduce an
.rtf
> > table holding my desired statistics.
> >
> > So far I have tried the following:
> > eststo psttest: pstest [varlist], sum t(d)
> > esttab pstest
> > That gave me the coeff. and t-stats from (what I think is) a probit on
> > treatment status on the matched sample.
> >
> > Additionally I have tried using [return] after pstest, my idea was to
> > combine it with estadd. But as far as I can tell [return] did not have
> > variable specific statistics.
> >
> > As the paper is still in version 1 and we balance on 46 variables and
> > examine outcomes for 3 different treatments stata generated tables is
> > considerable time-saving as well as error minimizing. Therefore, I will
be
> > thankful for all suggestions!
> >
> > Best,
> > Peer.
>
> *
> * 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/
*
* 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/
