[Carlo Lazzaro]

Barbara:
this result points you out to pooled OLS with standard errors clustered on -panelid-.

[Barbara Heerkens]

The thing is, when I want to cluster the standard errors on -panelid-
The F or chi2 model statistic has been reported as missing.
This only happens when I use -cluster(_j)- behind the regression, but _j is not nonzero for only one cluster/observation.

I also was not sure whether to include i._j in the regression itself or not, this does not matter for the F or chi2 model statistic, but makes the p-values veeery significant for the independent variables.... too significant to be true actually..

reg returns post##conn size lnde leverage i.sic2 i.date, robust cluster(_j)
note: 17596.date omitted because of collinearity

Linear regression                               Number of obs     =     20,174
                                                F(75, 76)         =          .
                                                Prob > F          =          .
                                                R-squared         =     0.1807
                                                Root MSE          =     .01363

                                      (Std. Err. adjusted for 77 clusters in _j)
--------------------------------------------------------------------------------
               |               Robust
       returns |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
---------------+----------------------------------------------------------------
        1.post |  -.0461615   .0047399    -9.74   0.000    -.0556018   -.0367212
   1.connected |    .000405   .0001781     2.27   0.026     .0000503    .0007597
               |
post#connected |
          1 1  |   -.001209   .0003477    -3.48   0.001    -.0019016   -.0005164
               |
          size |   .0002179   .0000542     4.02   0.000     .0001101    .0003258
          lnde |  -.0000461   .0000935    -0.49   0.623    -.0002323      .00014
      leverage |   .0001324   .0006202     0.21   0.832    -.0011028    .0013676
               |
          sic2 |
           15  |  -.0007595   .0002061    -3.69   0.000    -.0011699   -.0003491
           20  |  -.0005695   .0002024    -2.81   0.006    -.0009727   -.0001663
           27  |  -.0005679   .0003602    -1.58   0.119    -.0012853    .0001494
           32  |   -.001205   .0001316    -9.16   0.000    -.0014671   -.0009429
           35  |  -.0012511   .0004759    -2.63   0.010    -.0021989   -.0003033
           36  |  -.0019702   .0002438    -8.08   0.000    -.0024557   -.0014847
           42  |  -.0006698   .0003733    -1.79   0.077    -.0014132    .0000736

[Carlo Lazzaro]

Barbara:
for missing F-test, see: -help j_robustsingular- (the section on clustered standard errors).
I would not add -i._j-, as it probably overfits the model.

[Barbara Heerkens]

Carlo, thank you for all your help! Much appreciated!