Within R Squared in xtscc

[Nathalie Ruth]

Dear all,

Is within R squared in -xtscc- the same as R squared or Adjusted R squared? If not, what is the difference?
And how to interpret the within R squared result?

Thank you for your kind help.

[Maxence Morlet]

Within R squared will tell you how much of the variation in your dependent variable within one unit across time is explained by your model.

The overall R squared will tell you how much of the overall variation (so both between and within if you have panel data) in your dependent variable is explained by your model.

The adjusted R squared adds a penalty for weak regressors (i.e. regressors that take up a degree of freedom but do not contribute to explaining variations in Y).

[Nathalie Ruth]

Hi, thank you for your response.
-xtscc- only reports the within R squared. If my thesis required me to report the R squared and Adjusted R squared, from where do I get these results?
I'm using fixed effect model.

Thank you!

[Maxence Morlet]

So the generic command for the adjusted R squared is di e(r2_a) (postestimation)

I'm not sure of its relevance in this context however.

By the way, in your panel data do you have T>N? And is the number of time periods very large?

[Nathalie Ruth]

Hi, I have N>T, N = 77 T = 6.

[Maxence Morlet]

Ok so your data are not suited for xtscc unfortunately... xtscc relies on large T/N asymptotics, i.e. the ratio T/N must tend towards infinity.

If I were you I would use Correia's reghdfe and two way fixed effects with robust cluster standard errors