Re: st: RE: using xtivreg2 for FE

[Murod Aliyev <ma.statalist@yahoo.com>]

Thanks very much for your reply Eric!

Yes, the sample size is small indeed (it is actually a population of firms in 
one industry in one country). Using OLS would make it simpler, certainly. I just 
could not do the autocorrelation test with standard regress postestimation 
commands (dwatson/durbinalt/bgodfrey), as they are for time series data and do 
not allow panles: error message "sample may not include multiple panels". Is 
there any other test that would be appropriate?

newey command didn't work either, saying "year is not regularly spaced". xtregar 
does not test, but specify AR level, and I have no evidence to specify any level 
of serial correlation. Moreover, xtregar does not do anything with 
heteroskedasticity.

I was familiar with xtivreg2 before, and thought I might use it in this 
situation.

OLS estimates are of course the same as FE estimates. the only purpose of using 
xtivreg2 here is getting autocorrelation robust standard errors.

I will be happy to have any suggestions!

Best regards,
Murod





----- Original Message ----
From: DE SOUZA Eric <eric.de_souza@coleurope.eu>
To: "statalist@hsphsun2.harvard.edu" <statalist@hsphsun2.harvard.edu>
Sent: Mon, 21 February, 2011 12:37:24
Subject: RE: st: RE: using xtivreg2 for FE

Your sample is much too small to do anything sophisticated: even -fe robust-
With only seven firms I would introduce dummy variables for the firms and stick 
to OLS
Then check the residuals for mis-specification


Eric de Souza
College of Europe
Brugge (Bruges), Belgium
http://www.coleurope.eu


-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu 
[mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Murod Aliyev
Sent: 21 February 2011 12:15
To: statalist@hsphsun2.harvard.edu
Subject: Re: st: RE: using xtivreg2 for FE

Dear Mark,
Thanks very much for your reply!
Time span is  9 years in average (8-11 years unbalances). Is it too bad? Number 
fo firms is only 7 (so total N of obs =70).
I tried bw(), bw(2), bw(auto). Results vary, but not to a large extent.
Thanks again!
Murod





----- Original Message ----
From: "Schaffer, Mark E" <M.E.Schaffer@hw.ac.uk>
To: statalist@hsphsun2.harvard.edu
Sent: Sun, 20 February, 2011 19:20:54
Subject: st: RE: using xtivreg2 for FE

Murod,

> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu 
> [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of 
> Murod Aliyev
> Sent: 19 February 2011 14:04
> To: statalist@hsphsun2.harvard.edu
> Subject: st: using xtivreg2 for FE
> 
> Dear all,
> I want to estimate a simple FE model, but there may be 
> heteroskedasticity and autocorrelation. As far as I 
> understand xtreg fe robust gives Huber-White sandwich 
> estimator, which deals with heteroskedasticity only.
> Can I use xtivreg2 with fe robust bw() options to get HAC 
> (Het. and autocorrelation consistent) standard errors. Of 
> course, I will not specify any instruments, as I want only FE 
> estimation.

Yes, you can.

Two points worth noting: (a) -xtreg, fe robust- reports cluster-robust
SEs, not heteroskedasticity-robust SEs as you suggest.  (b) If you use
-xtivreg2- with the bw() option to get kernel-robust SEs, you are
relying on large-T asymptotics.  If your dataset is not very long in the
time-series dimension, the kernel-robust approach will not work very
well.

HTH,
Mark

> Thanks in advance!
> Murod
> 
> 
> 
>      
> 
> *
> *   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/



      

*
*   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/