Statalist The Stata Listserver


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

st: RE: How can I correct ivreg2 coefficients for AR1?


From   "Schaffer, Mark E" <M.E.Schaffer@hw.ac.uk>
To   <statalist@hsphsun2.harvard.edu>
Subject   st: RE: How can I correct ivreg2 coefficients for AR1?
Date   Wed, 13 Sep 2006 12:04:04 +0100

Adam,

> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu 
> [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of adam dvir
> Sent: 13 September 2006 08:10
> To: statalist@hsphsun2.harvard.edu
> Subject: st: How can I correct ivreg2 coefficients for AR1?
> 
> Dear all,
> 
> My aim is to run 2sls regression with correction to 
> auto-correlation AR(1).
> But I could not find a command that corrects AR(1) after 
> "ivreg" command (the same way "prais" command does to a 
> simple linear regression).

I think that's right - there isn't a built-in command.

> I heard that there is a user's command "ivreg2" that is able 
> to do so, but I am not sure I know how it works. I tried to 
> correct my regression with the
> bw(#) specification after ivreg2, but the coefficient did not 
> change, only the standart errors.

This is because what you've done is ask ivreg2 for standard errors that
are robust to arbitrary autocorrelation.  It's just like the usual
-robust- option - the standard errors become consistent, but the
coefficients don't change because they were already consistent (by
assumption), albeit possibly inefficient.

Note that this is a correction for arbitrary autocorrelation, so it's
more general than just your AR(1) case.

The bw(#) option specifies how quickly the autocorrelation is assumed to
die out.  The literature on this is pretty technical; what number you
choose depends among other things on the kernel estimator used and the
sample size.  One rule of thumb for the Barlett (a.k.a. Newey-West)
kernel is bandwidth = 0.75 * T^1/3 where T is the sample size.  See
Stock-Watson, Introduction to Econometrics, pp. 505-6.

> I tried the "gmm" specification, and this time the 
> coefficients did change, but I am not sure if the results are 
> IV coefficients or GMM's.

They are GMM coefficients.  The difference is that they are efficient in
the presence of arbitrary autocorrelation (not just consistent like the
IV coefficients).

Cheers,
Mark

> 
> Can you please direct me to the right answer?
> thank you very much
> 
> Adam Dvir
> 
> _________________________________________________________________
> All-in-one security and maintenance for your PC.  Get a free 
> 90-day trial!   
> http://www.windowsonecare.com/trial.aspx?sc_cid=msn_hotmail
> 
> *
> *   For searches and help try:
> *   http://www.stata.com/support/faqs/res/findit.html
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/
> 
> 

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   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   |   What's new   |   Site index