Re: st: Correction for Autocorrelation and Heteroskedasticity

[William Buchanan <william@williambuchanan.net>]

Hi June,

First, have you tried replicating this problem with any of the Stata datasets or with a dataset that you've generated on your own that you can provide to list members?  Second, please read the Statalist FAQ regarding the use of full names and providing enough detail to get a good response from the listserv members.

HTH,
Billy Buchanan



On Jan 7, 2013, at 7:11 AM, June wrote:

> Hi everyone,
> 
> I'm having trouble understanding what's going on when I correct for autocorrelation and heteroskedasticity in panel data.  From what I understand, these are issues that affect the standard errors, and not the point estimate, in a regression.  However, when I try to adjust for these issues using either xtgls or xtregar, my coefficients estimates change a lot, sometimes even changing signs! Can anyone explain why this would happen, or point me towards resources that would explain it?
> 
> For the record, I'm using stata 12, and comparing my results from:
> xtreg y x1 x2 x3, fe                   // no corrections
> xtgls y x1 x2 x3 i.groupvar, panels(i) corr(ar1) force // correct for ar1 autocorrelation
> xtgls y x1 x2 x3 i.groupvar, panels(hereto) corr(i) force        // correct for heteroskedastic errors
> xtgls y x1 x2 x3 i.groupvar, panels(hereto) corr(ar1) force    // correct for both
> xtregar y x1 x2 x3, fe rhotype(regress)     // correct for ar1 autocorrelation
> 
> Thank you!
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/faqs/resources/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/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/