Correlated Random Effects Goodness of Fit

[Caroline Schmidt]

Dear all,
I have a panel with 345 observations and six variables. My cross-sectional variable is Panel_bland (15 groups) and my time-series variable is Panel_year (23 years).
I xtreg a correlated random effect model (Mundlak). I am analysing the results and saw that no R-squared has been produced.
I know Stata must have a good reason for not displaying it as R-squared might not be a good measure.
What might be a good alternative to assess how much of the variation can be explained by the chosen variables?
Thank you in advance.
Caroline

[Carlo Lazzaro]

Caroline:
actually you can retrieve various R-squared from the community-contributed command -mundlak- (please note that you're kindly requested to declare when you use a non-official Stata command, for good reasons that are even better explained in the FAQ. Thanks), as you can se from the following toy-example:

. webuse nlswork, clear
(National Longitudinal Survey.  Young Women 14-26 years of age in 1968)

. mundlak ln_wage age south, stats(N N_g rho r2_o r2_w r2_b)

+------------------------------------------------+
|             Variable |     RE     |  Mundlak   |
|----------------------+------------+------------|
|                  age |      0.019 |      0.018 |
|                south |     -0.131 |     -0.077 |
|            mean__age |            |      0.006 |
|          mean__south |            |     -0.119 |
|                _cons |      1.169 |      1.049 |
|----------------------+------------+------------|
|                    N |      28502 |      28502 |
|                  N_g |   4710.000 |   4710.000 |
|                  rho |      0.583 |      0.583 |
|                 r2_o |      0.115 |      0.121 |
|                 r2_w |      0.104 |      0.104 |
|                 r2_b |      0.132 |      0.133 |
+------------------------------------------------+

That said, with such a large T dimension, I would have considered -xtgls- or -xtregar- instead of -xtreg-.

[Caroline Schmidt]

Dear Carlo,
thank you for your response and the suggestion.
I am not using the - mundlak - command as there were some specifics which required me to run the - xtreg...,re- and include the means manually.
Is there a way to compute the R-squared without using the mundlak package?
Thank you for your input on the t-dimensions and possible alternatives.
Kindly,
Caroline

[Carlo Lazzaro]

Caroline:
I'm not sure I'm following your approach right.
If you use -xtreg,re- it gives you back by default within, between (the most important) and overall R-squares.