Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down at the end of May, and its replacement, statalist.org is already up and running.


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

Re: R: st: intreg with control for sample selection and endogeneity bias


From   urbain thierry YOGO <yogout@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: R: st: intreg with control for sample selection and endogeneity bias
Date   Tue, 30 Oct 2012 18:08:24 +0100

To the best of my knowledge, boostrap is allowed as an option of
intreg. About the endogeneity, i guess you have a specific variable
that you suspect to be endogenous. You should find valid instrument
(strongly correlated with the endogenous variable and not correlated
with the error term). You can run a first step equation (endogenous
variable= f(insrument, exogenous variables) and recover the predicted
value that you can insert in the wage equation. The main issue i think
is how to compute standards errors. May be you can do that using
boostrap, but i am not sure. Another issue is that it become difficult
to ensure that your instrument is not weak (the Stock and Yogo 2005
test apply only in linear model to the best of my knowledge) neither
valid (apply the Hansen test). But you can check the same procedure
using OLS to see whether results change dramatically.
Thierry

2012/10/30, Justina Fischer <JAVFischer@gmx.de>:
> Hi Francesco
>
> obtaining bootstrapped s.e. should be an option of -intreg-.
>
>
> For the first stage: Practically a good model specification strategy
> (instrument, exclusion restriction!) with supportive statistical tests
> should do at least for the first stage.
>
> For the second stage you need to consult an intreg-specialist which I am
> not....
> E.g. I do not know whether the assummption of the joint normality of the
> error terms still prevails.
>
> HTH
>
> Justina
> -------- Original-Nachricht --------
>> Datum: Tue, 30 Oct 2012 00:14:08 +0100
>> Von: "Francesco Pastore" <francesco.pastore@unina2.it>
>> An: statalist@hsphsun2.harvard.edu
>> Betreff: R: st: intreg with control for sample selection and endogeneity
>> bias
>
>> Hi, Justina,
>> any more details about bootstrapping and the rest? How would you suggest
>> doing it practically.
>> Francesco
>>
>> -----Messaggio originale-----
>> Da: owner-statalist@hsphsun2.harvard.edu
>> [mailto:owner-statalist@hsphsun2.harvard.edu] Per conto di Justina
>> Fischer
>> Inviato: 30 October 2012 00:09
>> A: statalist@hsphsun2.harvard.edu
>> Oggetto: Re: st: intreg with control for sample selection and endogeneity
>> bias
>>
>> Hi Francesco
>>
>>
>> when pursuing Thierry's suggesion do not forget to get the standard
>> errors
>> of the intreg regression right; I think bootstrapping is the easiest.
>>
>> In general, make sure that the statistical assumptions of both stages are
>> met since they are important for identification and consistency.
>>
>> Justina
>>
>> -------- Original-Nachricht --------
>> > Datum: Mon, 29 Oct 2012 23:12:45 +0100
>> > Von: urbain thierry YOGO <yogout@gmail.com>
>> > An: statalist@hsphsun2.harvard.edu
>> > Betreff: Re: st: intreg with control for sample selection and
>> > endogeneity bias
>>
>> > hi
>> > i think you can deal with the sample selection as usual : you estimate
>> > an equation of labor market participation, you recover the predicted
>> > value p and you  compute the inverse Mills ratio as follow:
>> > gen pdfp1=normalden(p1)
>> > gen cdfp1=normprob(p1)
>> > gen IMR=pdfp1/cdfp1
>> > finally, you insert IMR in the wage equation.
>> >
>> > 2012/10/29, Francesco Pastore <francesco.pastore@unina2.it>:
>> > > Hi, All,
>> > > I would like to ask to the statalist the following question:
>> > >
>> > > I am estimating a wage equation to see the impact of overeducation
>> > > on earnings of a sample of graduates observed 5 years after getting
>> > > their degree. The dependent variable is defined as an interval data:
>> > > people
>> > are
>> > > asked to say to which class of earnings does their income belong to,
>> > rather
>> > > than a precise value. In this case, I should use the intreg command.
>> > > One problem is that these types of cross-section estimates are
>> > > plagued by sample selection and endogeneity bias. Does anybody know
>> > > how to deal with it in the contest of interval regressions? Is there
>> > > any routine available in
>> > Stata?
>> > >
>> > > Best regards
>> > > Francesco
>> > >
>> > >
>> > >
>> > > *
>> > > *   For searches and help try:
>> > > *   http://www.stata.com/help.cgi?search
>> > > *   http://www.stata.com/support/faqs/resources/statalist-faq/
>> > > *   http://www.ats.ucla.edu/stat/stata/
>> > >
>> >
>> >
>> > --
>> > *Urbain Thierry YOGO
>> > Ph.D candidate in Economics*
>> > *
>> > *   For searches and help try:
>> > *   http://www.stata.com/help.cgi?search
>> > *   http://www.stata.com/support/faqs/resources/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/faqs/resources/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/faqs/resources/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/faqs/resources/statalist-faq/
> *   http://www.ats.ucla.edu/stat/stata/
>


-- 
*Urbain Thierry YOGO
Ph.D candidate in Economics*
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/


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