Hi Statalist,

I used the following code
Code:
areg grade shock#c.months_exposure#c.treatment_intensity shock months_exposure treatment_intensity i.round,a(id) cluster (cluster_id)
and got the following output
Code:
------------------------------------------------------------------------------------------------------
                                     |               Robust
                            grade |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------------------------------+----------------------------------------------------------------
      shock#c.months_exposure#|
               c.treatment_intensity |
                                  0  |   3.06e-06   3.68e-06     0.83   0.416    -4.64e-06    .0000108
                                  1  |   1.34e-06   3.47e-06     0.39   0.704    -5.93e-06    8.60e-06
                                     |
                           shock |   .0428199   .0584205     0.73   0.473    -.0794556    .1650955
                     months_exposure |    .011039   .0155351     0.71   0.486    -.0214763    .0435543
                 treatment_intensity |  -.0001264     .00028    -0.45   0.657    -.0007124    .0004596
                                     |
                               round |
                                  1  |          0  (empty)
                                  2  |          0  (omitted)
                                  3  |   1.337061   .3766628     3.55   0.002     .5486966    2.125425
                                  4  |   4.615511   1.075007     4.29   0.000     2.365495    6.865527
                                  5  |    7.13307   1.617598     4.41   0.000     3.747399    10.51874
                                     |
                               _cons |   .1175795   .0889333     1.32   0.202    -.0685599    .3037189
-------------------------------------+----------------------------------------------------------------
                                  id |   absorbed                                    (1856 categories)
shock is a binary variable and treatment intensity and months of exposure are continuous variables.
As can be seen, the regression table reports coefficient when the interaction term is zero. Shouldn't the coefficient in that case be zero, and hence be omitted?
For instance, with this specification
Code:
areg grade shock##c.treatment_intensity i.round,a(id) cluster(cluster_id)
I get the result:
Code:
---------------------------------------------------------------------------------------------------------
                                        |               Robust
                               grade |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
----------------------------------------+----------------------------------------------------------------
                            1.shock |   .0320189   .0630621     0.51   0.617    -.0999716    .1640094
                    treatment_intensity |   .0000266   .0001911     0.14   0.891    -.0003734    .0004266
                                        |
shock#c.treatment_intensity |
                                     1  |  -.0000893   .0001444    -0.62   0.544    -.0003915    .0002129
                                        |
                                  round |
                                     1  |          0  (empty)
                                     2  |  -5.459317   .1785494   -30.58   0.000    -5.833026   -5.085609
                                     3  |  -3.851622   .0677831   -56.82   0.000    -3.993493    -3.70975
                                     4  |          0  (omitted)
                                     5  |   2.898968   .0800272    36.22   0.000     2.731469    3.066467
                                        |
                                  _cons |   5.584258   .1059441    52.71   0.000     5.362515    5.806002
----------------------------------------+----------------------------------------------------------------
                                     id |   absorbed                                    (1856 categories)
Here the coefficient for zero interaction is omitted.

Can anyone please explain why this difference in reporting coefficients? I feel I'm missing something simple but I'm unable to figure it out.

Thanks,