st: RE: R2 difference between PCSE & DKSE/ OLS regression models

["Jacobs, David" <jacobs.184@sociology.osu.edu>]

In -xtreg, fe- the within R2 does not take the case dummy-intercept effects into account. I am less sure how the R2 is calculated in -xtpcse- but I think it differs from the -regress- R2.

Dave Jacobs

-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Petya Alexandrova
Sent: Friday, January 18, 2013 11:17 AM
To: statalist@hsphsun2.harvard.edu
Subject: st: R2 difference between PCSE & DKSE/ OLS regression models

Dear Statalist users,
  
I am estimating pooled time-series cross-sectional models in stata. I first run a simple OLS regression with fixed effects. Since some assumptions were not met (presence of heteroskedasticity, cross-sectional correlation and possibly also serial correlation) i decided to run models which can control for the problems. I went for the Driscoll and Kraay st. errors (stata command xtscc) and panel-corrected st. errors (stata command xtpcse) models, keeping the fixed effects in both cases. The R2 for the OLS regression is almost the same as for the one with Driscoll and Kraay st. errors, appr. 0.10. However, the model with panel-corrected st. errors gets an R2 of 0.65. Is this difference normal? What can it be explained by?
  
Here is the summary of the commands for the three models:
  1) xtreg y x z w laggedy, fe
  2) xtscc y x z w laggedy, fe
  3) xtpcse y x z w laggedy d1 d2 d3 d4 d5 d6, het corr(ar1)
  where d1... d6 stand for the respective dummies
  
Thank you!
  
Best,
  Petya Alexandrova
  
*
*   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/