R squared and Adjusted R squared values on xtreg (random effects panel regression)

[Cen Hua]

Hello everyone!

I am doing my dissertation which requires me to do some random effects panel data regression.

I need to prove/disprove that adding a certain independent variable to the model improves it. I plan to run both regressions (one without the new variable one that includes the new variable) and compare their adjusted R squared values. But from some previous posts here, I found that random effects do not compute the adjusted R squared.

I am using Stata/IC 15.1, is there a way to acquire the adjusted R squared? Otherwise, what alternative measurements apart from adjusted R squared can help me in determining if those added independent variables improve the model?

[Carlo Lazzaro]

Cen:
welcome to this forum.
See: https://www.statalist.org/forums/for...r2-in-xtreg-re

[Christopher Bratt]

The link to an earlier Statalist post is broken.

[Carlo Lazzaro]

Sorry for the mishap:
The following one shoud work: https://www.statalist.org/forums/for...-effects-model

[Cen Hua]

Hi Carlo

I have tried this code you recommended:
ereturn list di e(r2_a) But it shows no adjusted r squared value (it works for fixed effects but not random effects)


And in that forum post, I'd like to re-clarify Barbara's interpretation of your message: "you rather look at the between R-sq instead of the adjusted R sq". Does this mean this new variable improves the model if the between R squared is higher in the model with the new variable?

Also, what would be the reason for this (to better look at the between R squared than the adjusted R squared)? I need to be able to justify using the between R squared as an alternative to adjusted R squared.

Thanks in advance!

[Carlo Lazzaro]

Cen:
please see my reply #6 under the same thread.
As far as your second question is concerned, the core meaning is that between R-sq is more meaningful than overall R-sq when you run -xtreg,re-.