Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.


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

RE: st: can we include dummy variables in xtpcse/xtgls models


From   Lopa Chakraborti <[email protected]>
To   "[email protected]" <[email protected]>
Subject   RE: st: can we include dummy variables in xtpcse/xtgls models
Date   Fri, 11 Mar 2011 11:29:05 -0500

Felix,
thanks much for your inputs. I will check for heteroskedasticity in my sample. 

As for the models, I am not sure I understand the restriction you have identified in terms of number of panels- is it 11 panels that you are referring to. My sample has 100 panels and a time units varying from 4 to 12 per panel. 

Also, I am trying to figure out why the estimated autocorrelation coefficient changes sign from positive (around 0.7) to negative (around -0.2) upon including the panel level dummy variables. Is xtpcse/xtgls (with panel dummies) not appropriate because my sample has too many panels and too few observations per panel. The estimated autocorrelation coefficient using xtregar with FE option is however positive (0.7).
many thanks
Lopa    
________________________________________
From: [email protected] [[email protected]] On Behalf Of Felix Wädlich [[email protected]]
Sent: Friday, March 11, 2011 11:07 AM
To: [email protected]
Subject: Re: st: can we include dummy variables in xtpcse/xtgls models

Dear Lopa,

really not an expert, but I guess i can give u a technical answer:
u can implement FE in -xtpcse- simply by adding individual dummies
through putting an -i.- before the variable that identifies ur group
variable, like -i.countrycode-. Though, this is only possible up until
11.000 units. So if u have a huge survey panel it wont work.
As I understand, -xtregar- will give u different estimates, simply
because -xtpcse, c(ar1)- not only considers an AR1 autocorrelation
structure but also heteroskedasticity and contemporaneously
correlation, which are common problems in panels, with long times
series. For the latter ones u can test for with -xttest3- and -xtcsd-
(which u have to install, since there are not Stata build-ins).

Best,
Felix

2011/3/11 Lopa Chakraborti <[email protected]>
>
> Dear Statalist,
> I wanted to find out if one can include dummy variables in xtpcse/xtgls models in order to capture 'fixed effects' of the panel variable. And, would this be similar to estimating xtregar model with FE option but with the possibility of including time invariant regressors.
> many thanks in advance for any advice
> Lopa
> *
> *   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 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index