R² random panel data
Posted: Tue Dec 17, 2013 10:49 am
Hi,
We have run panel data regressions with random effects in the Eviews and Stata. After analyzing the results, we have noted that output of both models were considerably different.
Eviews provided us with weighted an unweighted R-square results while Stata gave us an overall R-square, without mentioning if the R-square is a weighted or unweighted.
We understand that panel data regression with random effects requires the correction of estimator variance and therefore the right output to consider is the weighted R-square. In order to run the models, we use Xtreg variables, re Stata command.
In sum, we have 3 questions:
1- What type of R-square value (unweighted or weighted) does Stata provide in case of panel data regressions with random effects?
2- Is there any Stata command that would provide outputs with unweighted or weighted R-square results?
3- Would you recommend the use of unweighted or weighted R-square results panel data regressions with random effects ? Which of both concepts better approach the predictive capacity of the regression model?
Tks,
We have run panel data regressions with random effects in the Eviews and Stata. After analyzing the results, we have noted that output of both models were considerably different.
Eviews provided us with weighted an unweighted R-square results while Stata gave us an overall R-square, without mentioning if the R-square is a weighted or unweighted.
We understand that panel data regression with random effects requires the correction of estimator variance and therefore the right output to consider is the weighted R-square. In order to run the models, we use Xtreg variables, re Stata command.
In sum, we have 3 questions:
1- What type of R-square value (unweighted or weighted) does Stata provide in case of panel data regressions with random effects?
2- Is there any Stata command that would provide outputs with unweighted or weighted R-square results?
3- Would you recommend the use of unweighted or weighted R-square results panel data regressions with random effects ? Which of both concepts better approach the predictive capacity of the regression model?
Tks,