RE: st: RE: ivreg2 post-estimation tests

["Schaffer, Mark E" <M.E.Schaffer@hw.ac.uk>]

Gordon,

> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu 
> [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of 
> Abekah Nkrumah
> Sent: 25 May 2012 20:36
> To: statalist@hsphsun2.harvard.edu
> Subject: Re: st: RE: ivreg2 post-estimation tests
> 
> Dear Mark,
> 
> Thank you very much for the reply. I should have included in 
> my mail that am using the cluster robust covariance matrix as 
> well due to intra-cluster correlations. I had earlier tried 
> the option you have suggested but the response from stata was 
> that overid is not valid when using a cluster robust 
> covariance matrix. Thus the decision to consider the ivreg2 
> since I have read lots of it literature and know that the 
> post-estimation and first-stage statistics are valid even in 
> the presence of a cluster robust covariance matrix.
> 
> So am some how at a stand still. Thanks very much and will 
> wait for your feedback
> 
> 
> Gordon
> 
> 
> 
> On Fri, May 25, 2012 at 4:40 PM, Schaffer, Mark E 
> <M.E.Schaffer@hw.ac.uk> wrote:
> > Gordon,
> >
> >> -----Original Message-----
> >> From: owner-statalist@hsphsun2.harvard.edu
> >> [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Abekah 
> >> Nkrumah
> >> Sent: 25 May 2012 14:22
> >> To: statalist@hsphsun2.harvard.edu
> >> Subject: st: ivreg2 post-estimation tests
> >>
> >> Dear All,
> >>
> >> I am running ivprobit with factor variables (eg. i.var_name).
> >> However, since the post estimation overid command does not support 
> >> factor variables, it means I will not be able to conduct my 
> >> overidentification test.
> >
> > Why don't you create your factor variables by hand and then use 
> > -ivprobit-?  That is, first generate the factor variables 
> so that they 
> > are standard Stata variables, and then run your IV probit.  
> Then you 
> > can use the overidentification test for IV probit provided 
> by -overid-.
> >
> > --Mark
> >
> >> Thinking through this, I
> >> have decided to run a linear probability model using ivreg2 since 
> >> that support factor variables and therefore will be able 
> to have the 
> >> test results for overidentification test.
> >>
> >> Now considering that the linear probability model is not that 
> >> different from the probit model, will the test results for
> >> (i) overidentification, (ii) underidentification (iii) F test for the 
> >> joint significance of the instruments (iv) Stock Yogo calculations, 
> >> be valid for my ivprobit estimation?

I think the answer is to (i) is "sort of"; the answer to (ii) and (iii) (which are the same thing) is "not really but better than nothing"; and the answer to (iv) is "no".

But maybe there is a simpler solution to your problem.  You should be able to use -gmm- to estimate your IV probit, and allow for within-group correlation by using the cluster-robust VCE.  -estat overid- after estimation with -gmm- would give you a cluster-robust overidentification test statistic.  Estimating the first stage by hand using -regress- with a cluster-robust VCE should (I think) let you test for underidentification in the usual way (but not weak identification - AFAIK, that's not been worked out yet).  And if you've not used -gmm- before, there is a probit example in the -gmm- help file to get you started.

Cheers,
Mark

> >>
> >> I will appreciate some help on this. Thank you very much
> >>
> >> Regards
> >>
> >> Gordon
> >> *
> >> *   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/
> >>
> >
> >
> > --
> > Heriot-Watt University is the Sunday Times Scottish 
> University of the 
> > Year 2011-2012
> >
> > Heriot-Watt University is a Scottish charity registered 
> under charity 
> > number SC000278.
> >
> >
> > *
> > *   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/
> 


-- 
Heriot-Watt University is the Sunday Times
Scottish University of the Year 2011-2012

Heriot-Watt University is a Scottish charity
registered under charity number SC000278.


*
*   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/