John,
> -----Original Message-----
> From: [email protected]
> [mailto:[email protected]] On Behalf Of
> John Antonakis
> Sent: Thursday, September 10, 2009 2:17 PM
> To: [email protected]
> Subject: Re: st: 3SLS / Three Stage with heteroscedasticity
> robust errors
>
> Hi Mark:
>
> The help file mentions too:
>
> cmp is appropriate for two types of models: 1) those in which a truly
> recursive data-generating process is posited;
> and 2) those in which there is simultaneity, but
> instruments allow
> the construction of a recursive set of equations,
> as in two-stage least squares, that can be used to consistently
> estimate structural parameters in the final stage. In
> the first case, cmp is a full-information maximum
> likelihood (FIML)
> estimator, and all estimated parameters are
> structural. In the latter, it is a limited-information (LIML)
> estimator, and only the final stage's coefficients are
> structural, the rest being reduced-form parameters. What
> matters for
> the validity of cmp is that the system of
> equations is recursive, whether or not the model is.
>
> Note, the presentation of Roodman (on page 12):
>
> http://www.stata.com/meeting/dcconf09/dc09_roodman.ppt
>
> What do you make of this?
I think it's the following, but maybe I'm missing something.
You can use -cmp- for a genuinely recursive model, i.e., the
fully-specified model is indeed recursive. This is full-information,
system estimation - you're estimating every parameter there is to
estimate.
You can use -cmp- to estimate part of a fully-specified model, if you
can write it out in a set of recursive equations in such a way that the
model parameters you're interested in are in the part being estimated.
The IV/2SLS example would say it's a limited information estimation of
part of a model if the parameters of first-stage equations have no model
interpretation and are just mechanical byproducts of linear projection.
And it's a full information estimation of the full model if the
parameters of the first-stage equations are actually part of the model,
with behaviour interpretations etc. etc. In both cases, the equations
are recursive, but only in the latter case is the model recursive.
That's my take, at any rate.
Cheers,
Mark
> Best,
> J.
>
>
> ____________________________________________________
>
> Prof. John Antonakis
> Associate Dean Faculty of Business and Economics
> University of Lausanne
> Internef #618
> CH-1015 Lausanne-Dorigny
> Switzerland
>
> Tel ++41 (0)21 692-3438
> Fax ++41 (0)21 692-3305
>
> Faculty page:
> http://www.hec.unil.ch/people/jantonakis&cl=en
>
> Personal page:
> http://www.hec.unil.ch/jantonakis
> ____________________________________________________
>
>
>
> On 09.09.2009 18:00, Schaffer, Mark E wrote:
> > Wiebke, John,
> >
> >
> >> -----Original Message-----
> >> From: [email protected]
> >> [mailto:[email protected]] On Behalf Of
> >> John Antonakis
> >> Sent: Wednesday, September 09, 2009 4:52 PM
> >> To: [email protected]
> >> Subject: Re: st: 3SLS / Three Stage with heteroscedasticity
> >> robust errors
> >>
> >> Hi:
> >>
> >> Use Roodman's cmp. It allows you to estimate systems of
> >> equations using
> >> ML with heteroskedastic robust SEs. Do -ssc install cmp-.
> >>
> >
> > Good suggestion, and hopefully Wiebke's problem is
> recursive. From the
> > -cmp- help file:
> >
> > "Recursive" means, however, that cmp can only fit sets of
> equations with
> > clearly defined stages, not ones with simultaneous
> causation. A and B
> > can be modeled determinants of C and C a determinant of
> D--but D cannot
> > be a modeled determinant of A, B, or C.
> >
> > Cross fingers!
> >
> > --Mark
> >
> >
> >> HTH,
> >> J.
> >>
> >> Roodman, D. (2008). Cmp: Stata module to implement conditional
> >> (recursive) mixed process estimator.
> >> http://ideas.repec.org/c/boc/bocode/s456882.html.
> >>
> >> ____________________________________________________
> >>
> >> Prof. John Antonakis
> >> Associate Dean Faculty of Business and Economics
> >> University of Lausanne
> >> Internef #618
> >> CH-1015 Lausanne-Dorigny
> >> Switzerland
> >>
> >> Tel ++41 (0)21 692-3438
> >> Fax ++41 (0)21 692-3305
> >>
> >> Faculty page:
> >> http://www.hec.unil.ch/people/jantonakis&cl=en
> >>
> >> Personal page:
> >> http://www.hec.unil.ch/jantonakis
> >> ____________________________________________________
> >>
> >>
> >>
> >> On 09.09.2009 17:47, Wiebke Schlanbusch wrote:
> >>
> >>> Dear Statalist-ers,
> >>>
> >>> for my diploma thesis I want to estimate a simultaneous
> >>>
> >> equations model with different data samples using -reg3-. I
> >> tested all my single equations on heteroscedasticity using
> >> -ivreg2- for the equations and -ivhettest- for the test. Some
> >> of my equations turn out to be heteroscedastic and other
> >> not... Therefore I would like to estimate the 3SLS models
> >> using robust errors, which is not implemented. Someone
> >> suggested to programm it in Stata, but unfortunately my
> >> programming skills are not worth mentioning nor do I have the
> >> programming manual. This is why I am asking if anyone ever
> >> did this "robust errors three stage"-programming and might be
> >> so kind to help me with the code.
> >>
> >>> (I am currently using Stata 10.1)
> >>>
> >>> Kind regards,
> >>> Wiebke Schlanbusch
> >>>
> >>>
> >>>
> >>>
> >>> *
> >>> * 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/
>
--
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/
