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 on April 23, and its replacement, statalist.org is already up and running.


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

Re: st: about ivregress


From   John Antonakis <John.Antonakis@unil.ch>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: about ivregress
Date   Wed, 29 Aug 2012 20:42:47 +0200

One more thing, Lynn--this discussion has come up a couple of times; 2sls (i.e., the ivreg-type estimators) usually include all exogenous controls in all equations by default. Thus, to be on safer statistical ground, you would need to include x1 in all equations:

reg3 (y x1 x2 x5 ) (x2 x3 x4 x1) (x5 x6 x7 x1), 2sls

Under some circumstances you'd be fine putting x1 in the y equation, but not in the x2 and x5 equation. Though it is always safer to put x1 in all equations. For details see:

Baltagi, B. H. (2002). Econometrics. New York: Springer.

See also: http://www.stata.com/support/faqs/statistics/instrumental-variables-regression/

Best,
J.

__________________________________________

Prof. John Antonakis
Faculty of Business and Economics
Department of Organizational Behavior
University of Lausanne
Internef #618
CH-1015 Lausanne-Dorigny
Switzerland
Tel ++41 (0)21 692-3438
Fax ++41 (0)21 692-3305
http://www.hec.unil.ch/people/jantonakis

Associate Editor
The Leadership Quarterly
__________________________________________

On 29.08.2012 18:59, John Antonakis wrote:
Welcome.

Though I must have been very tired that day because the second sem code was incorrect, even after my second mea culpa. The first sem code, with ", covstructure(e._OEn)" was right. For the second one, if you wish to explicit correlate the disturbances you should do ", covstructure(e.x2*e.x5 e.y*x2 e.y*x5)". The Wald postestimation test of the covariance = 0 would be the Hausman test.

Best,
J.

__________________________________________

Prof. John Antonakis
Faculty of Business and Economics
Department of Organizational Behavior
University of Lausanne
Internef #618
CH-1015 Lausanne-Dorigny
Switzerland
Tel ++41 (0)21 692-3438
Fax ++41 (0)21 692-3305
http://www.hec.unil.ch/people/jantonakis

Associate Editor
The Leadership Quarterly
__________________________________________

On 29.08.2012 02:29, Lynn Lee wrote:
>
> Thank you, Prof. Antonakis. My model has two endogenous variables and each
> has its own instrumental variables. Your suggestion really solves my
> question.
>
> At same time,  thank Kit and Justina for kind reminder.
>
> Best Regards,
> Lynn Lee
>
> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu
> [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of John Antonakis
> Sent: Tuesday, August 28, 2012 1:14 AM
> To: statalist@hsphsun2.harvard.edu
> Subject: Re: st: about ivregress
>
> [Double Oops. Sorry about this.....I just realized that I messed up with the
> sem command too; the disturbances must be correlated. Here goes again.
> Apologies for my oversight:
>
>
> 1. Oops one: There was a typo in the Three-stage least square command, which > should have 3sls (the default, thus it does not even need to be written,
> instead of 2sls)
>
> 2. Oops two: in the sem command, you must correlate cross-equation
> disturbances; else, the estimator will not be an instrumental variable one.]
>
> It does; as Kit suggested "varlist" is a list of (n) variables. Try:
>
> ivregress 2sls y x1 (x2 x5 =x3 x4 x6 x7)
>
> If you want to specify your model precisely as indicated below, you'll need
> to use reg3 or sem:
>
> Two-stage least squares estimator
> reg3 (y x1 x2 x5) (x2 x3 x4) (x5 x6 x7), 2sls
>
> Three-stage least squares estimator
> reg3 (y x1 x2 x5) (x2 x3 x4) (x5 x6 x7), 3sls
>
> You can test overidentifying restrictions with the userwritten -overid-
> command.
>
> You can estimated this with -sem- too, Maximum likelihood:
>
> sem (y<- x1 x2 x5) (x2 <-x3 x4) (x5<-x6 x7), covstructure(e._OEn)
>
> or more specifically:
>
> sem (y<- x1 x2 x5) (x2 <-x3 x4) (x5<-x6 x7), covstructure(e.x2*e.x5)
>
> The covariance e.x2*ex5 is the Hausman test (see: Antonakis, J., Bendahan,
> S., Jacquart, P., & Lalive, R. (2010). On making causal
> claims: A review and recommendations. The Leadership Quarterly, 21(6),
> 1086-1120)
>
> You can also add ", vce(robust)" to the sem command.
>
> This should help,
> J.
>
> __________________________________________
>
> Prof. John Antonakis
> Faculty of Business and Economics
> Department of Organizational Behavior
> University of Lausanne
> Internef #618
> CH-1015 Lausanne-Dorigny
> Switzerland
> Tel ++41 (0)21 692-3438
> Fax ++41 (0)21 692-3305
> http://www.hec.unil.ch/people/jantonakis
>
> Associate Editor
> The Leadership Quarterly
> __________________________________________
>
> On 28.08.2012 19:05, Lynn Lee wrote:
>>
>> The example provided in help document is " ivregress 2sls rent
>> pcturban (hsngval = faminc i.region)", which indicates the endogenous
>> variable is "hsngval" and the number of endogenous variable is one.
>> Now, my model has two endogenous variables, "ivregress 2sls y x1
>> (x2=x3 x4) (x5=x6
> x7)", this
>> command is wrong. I am wondering whether there is command that
> handles two
>> endogenous variables x2 and x5 at same time. Thanks for any suggestion.
>>
>> Best Regards,
>> Lynn Lee
>>
>> -----Original Message-----
>> From: owner-statalist@hsphsun2.harvard.edu
>> [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of
> Christopher Baum
>> Sent: Monday, August 27, 2012 1:41 PM
>> To: statalist@hsphsun2.harvard.edu
>> Subject: re: st: about ivregress
>>
>> <>
>> Lynn said:
>>
>> My estimation model has two endogenous explanatory variables
> (cross-section
>> data set). The command "ivregress 2sls" or "ivregress gmm" are
> suitable for
>> one endogenous variable. What specific command I can use if I want to
> handle
>> these two endogenous variables at same time in Stata? And what
> command for
>> handling more than one endogenous variables in panel data is used?
>>
>>
>> The syntax for -ivregress- is
>>
>>
>>         ivregress estimator depvar [varlist1] (varlist2 = varlist_iv)
> [if]
>> [in] [weight] [, options]
>>
>> Note that -varlist2- is a varlist, not a varname. Thus you may put as
> many
>> variables as you wish into -varlist2-, s.t. the order condition for
>> identification, which requires that -varlist_iv- has at least as many
>> elements as -varlist2-.
>>
>> In panel data, use -xtivreg-, which uses the same syntax.
>>
>> For a broader set of features, -ssc desc ivreg2- and -ssc xtivreg2-
> and the
>> two Stata Journal papers by Baum, Schaffer, Stillman referenced in
>> their help files.
>>
>> Kit
>>
>> Kit Baum   |   Boston College Economics & DIW Berlin |
>> http://ideas.repec.org/e/pba1.html
>>                              An Introduction to Stata Programming  |
>> http://www.stata-press.com/books/isp.html
>>   An Introduction to Modern Econometrics Using Stata  |
>> http://www.stata-press.com/books/imeus.html
>>
>>
>> *
>> *   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/
> *
> *   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/

*
*   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/


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