Statalist The Stata Listserver


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

Re: st: Re: xtabond2 - what does large N mean?


From   "Rodrigo A. Alfaro" <raalfaroa@gmail.com>
To   <statalist@hsphsun2.harvard.edu>
Subject   Re: st: Re: xtabond2 - what does large N mean?
Date   Thu, 10 May 2007 17:30:13 -0400

In a very simple way, AB is a kind of instrumental variable estimator (after taking orthogonal deviation, it looks more close to an average of 2SLS estimator ~ 2SLS estimator). From the linear moment restrictions that are (T-1)(T-2)/2 you have like "instruments" K = (T-1)(T-2)/2 then for T=15 K=91 (~100). In other words, if you have T=15 and N=100 I think that you can still use AB... but you may check the results using Within Group estimator and applying the correction suggested by Hahn and Kuersteiner (2002). Several details for AB 1-step are in Alvarez-Arellano (2003) paper, the tables in the appendix could help you with the bias. R

----- Original Message ----- From: "tutor um" <tutor2005um@googlemail.com>
To: <statalist@hsphsun2.harvard.edu>
Sent: Thursday, May 10, 2007 11:58 AM
Subject: Re: st: Re: xtabond2 - what does large N mean?



Thank you Rodrigo for your quick answer! That helped.

Another question concerning the size of N. If N=100 and xtabond uses
e.g. 100 instruments, would this still yield ok results? where would
be the limit?
Any help is appreciated.

Thank you.


On 5/9/07, Rodrigo A. Alfaro <raalfaroa@gmail.com> wrote:
Andreas,

N means number of groups, countries in your case. Your N (160) is fine for
few periods (T), maybe less than 8. You could check that the AB estimator is
not appropriate for T large. See 2 econometrica papers: Hahn-Kuersteiner
(2002) and Alvarez-Arellano (2003). Likely to find the working papers on the
web.

Rodrigo.


----- Original Message -----
From: "tutor um" <tutor2005um@googlemail.com>
To: <statalist@hsphsun2.harvard.edu>
Sent: Wednesday, May 09, 2007 2:03 PM
Subject: st: xtabond2 - what does large N mean?


> Hello everybody,
>
> I am trying to use xtabond2 for a panel dataset of around 160 > countries.
>
> In the paper "How To Do xtabond2"
> (http://repec.org/nasug2006/howtodoxtabond2.cgdev.pdf) Roodman on page
> 1 (line 2) refers ro large N panels.
>
> Quote: "Both are general estimators designed for situations with 1)
> "small T, large N" panels,
> meaning few time periods and many individuals; ..."
>
> I am wondering what exactly is meant by N?
>
> Is it the number of observations in the panel (in my case around 800)?
> Or is it the number of Groups (in my case the countries) in the panel?
>
>
> Also in what respect is N meant to be compared?
>
> xtabond2 warns if the number of instruments is too big. Does that in
> reverse mean everything is fine if it doesn´t?
> Are the results not suitable for interpretation at all if a warning
> has been displayed while estimating? Or is it a hint for extra caution
> when interpreting results since they are displayed anyways.
>
> I hope someone can help me with my questions.
> Thank you so much for your help!
>
> Andreas Marns
>
> *
> * 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/

*
* 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