Hi Pallab,
I don't remember the default option for -xtregar-, but it sounds like the
initial observation is used as the same as Prais procedure for
autocorrelation. See the manual or any econometric textbook to see how the
initial observation is used. You will also note that the autocorrelation
issue is removed using semi-differences which imply that exogenous variables
are lagged as well.
Rodrigo.
----- Original Message -----
From: "Pallab Mozumder" <[email protected]>
To: <[email protected]>
Sent: Sunday, December 24, 2006 1:03 PM
Subject: st: number of observations in xtregar
Dear STATA Listers,
Regards. I am running a panel data model in which I
have country as the panel variable and then I have
country specific data for three years (I use tsset
country year). First I use the Xtreg with the default
option (which is random effects) and I get the
results. However, as I suspect the possibility of some
serial correlation, I would like to try this panel
regression with autoregressive error structure. So I
run Xtregar with the default option (which is also
random effects) which, as you know, allows AR(1)
process in the error structure. I get the results but
the number of observations that are used are exactly
same in both cases. Isn't it obvious that two of the
three periods would be used in the later case
(Xtregar) since the AR would need to use one of the
periods as a baseline. I checked these two
specifications with another dataset and the same thing
happens. Can anybody tell me why this is happening?
Thank you so much. Pallab
__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com
*
* 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/
