Statalist The Stata Listserver

[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

[no subject]

From   "Stephan Brunow" <>
To   <>
Date   Thu, 10 May 2007 21:58:54 -0400 (EDT)>
In-Reply-To: <>
Subject: AW: st: 2SLS and Probit
Date: Thu, 10 May 2007 21:16:58 +0200
Organization: =3D?iso-8859-1?Q?Technische_Universit=3DE4t_Dresden?=3D
Message-ID: <000801c79337$c87b8110$59728330$>
MIME-Version: 1.0
Content-Type: text/plain;
X-Mailer: Microsoft Office Outlook 12.0
Thread-Index: AceTJEdwIyzhm25CTPu1I3iTvGaHDQAE3wLA
Content-Language: de
Content-Transfer-Encoding: 8bit
X-MIME-Autoconverted: from quoted-printable to 8bit by
  id l4AJHCJ27763
Precedence: bulk
Austin, of course. There is that particular warning! However, if I run
exactly the steps of that faq with "hand made" regression, these structural
equation results are exactly the same as if I do it with -reg3- and the
option 2SLS (as long as the Y-variables are continuous). My, or better the
students question is: What is with dummy coded Y-variables in a system of
equations. How could he do the correction of the standard errors...

-----Urspr=FCngliche Nachricht-----
[] Im Auftrag von Austin Nichols
Gesendet: Donnerstag, 10. Mai 2007 18:57
Betreff: Re: st: 2SLS and Probit

Stephan Brunow --
Note that the FAQ you cite contains the "Warning: Instrumental
variables are commonly used to fit simultaneous systems models. What
follows is not appropriate for such models." However, -ivreg- (or
better, -ivreg2- avail via -ssc inst ivreg2, replace-) and -treatreg-
are both good options for estimating a regression of continuous Y on
endogenous indicator variable X using instruments Z, though they may
give quite different answers.

ssc inst ivreg2, replace
ssc inst esta, replace
reg lw80 lw s expr mrt
est sto ols
ivreg2 lw80 lw s expr (mrt=3Dmed iq age)
est sto iv
treatreg lw80 lw s expr, treat(mrt=3Dmed iq age lw s expr)
est sto treat
esta *, mti keep(mrt) lab se

On 5/10/07, Stephan Brunow wrote:
> one of our students has the problem to run a simultaneous equation system
> (SEM) where the endogenous variables are discrete (0/1-outcomes).
> One suggested method to solve that problem is =96 as usual =96 2SLS. To o=
> knowledge there is no command in STATA doing this so that we have to
> it on our own. However, as in the continue variable case, the standard
> errors have to be corrected.
> Does the following procedure of standard error adjustment also applies to
> 2SLS-probit models?

*   For searches and help try:

*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index