Hi Andrei,
I don't think you should be concerned. Your model without the fixed effects
is telling you that first order autoregressive serial correlation is
relatively muted and with the fixed effects your best guess is that there is
essentially no serial correlation. This second result isn't 'bad', for
example, it is what you would get if you simulated some panel data that
really had no serial correlation in the errors. You might find it useful at
this point to investigate some of the tests out there for higher order
autoregressive serial correlation (Wooldridge 2006 and Greene both have
some).
Iain
On Thu, Apr 29, 2010 at 5:43 AM, Andrei Roman <acromanya at gmail.com> wrote:
Dear all,
I am having trouble interpreting the Rho statistic for a panel-corrected
Prais-Winsten statistic with AR1. When fixed effects are introduced in the
regression, the rho statistic we are looking at changes from 0.110 to
-0.001. Some of the significant coefficients in the regression also change
dramatically. Should we be concerned about an rho statistic so close to
zero? Is there an optimal rho for this kind of model?
Thank you very much!
Andrei
_______________________________________________
gov2001-l mailing list
gov2001-l at
lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l