csdid : csdid does not return entire group ATTs when I include covariates

[Nathan Choi]

when I simply run csdid,

qui csdid y_var , ivar(statefip) time(t_ym) gvar(pol_ym) notyet long2
csdid_estat group


the results return all 10 groups.


. csdid_estat group
ATT by group
------------------------------------------------------------------------------
| Coefficient Std. err. z P>|z| [95% conf. interval]
-------------+----------------------------------------------------------------
GAverage | 39.75624 1.970523 20.18 0.000 35.89408 43.61839
G534 | -3.181626 3.323437 -0.96 0.338 -9.695444 3.332192
G588 | 11.59253 4.209699 2.75 0.006 3.341677 19.84339
G696 | 28.24867 5.484564 5.15 0.000 17.49912 38.99822
G702 | -13.40629 8.040015 -1.67 0.095 -29.16443 2.351854
G720 | 31.85622 16.74251 1.90 0.057 -.958487 64.67093
G726 | -26.01383 16.64317 -1.56 0.118 -58.63385 6.606177
G732 | 43.15997 15.62937 2.76 0.006 12.52696 73.79298
G744 | 19.30807 1.52619 12.65 0.000 16.31679 22.29935
G764 | 266.2424 6.36e-14 4.2e+15 0.000 266.2424 266.2424
------------------------------------------------------------------------------


However, when I include covariates, it returns some of the groups and the number of groups got returned differs by which covariates I include.

For example, when I include cov1
qui csdid y_var cov1 , ivar(statefip) time(t_ym) gvar(pol_ym) notyet long2
csdid_estat group
returns only 5 groups

ATT by group
------------------------------------------------------------------------------
| Coefficient Std. err. z P>|z| [95% conf. interval]
-------------+----------------------------------------------------------------
GAverage | .2577446 5.208761 0.05 0.961 -9.95124 10.46673
G534 | -6.54916 3.593409 -1.82 0.068 -13.59211 .4937928
G588 | 10.75291 1.911706 5.62 0.000 7.006041 14.49979
G696 | 15.31813 2.476017 6.19 0.000 10.46523 20.17104
G702 | -11.3719 5.898125 -1.93 0.054 -22.93201 .1882136
G720 | -6.861267 21.82422 -0.31 0.753 -49.63596 35.91342
------------------------------------------------------------------------------


and when I include cov2
qui csdid y_var cov2 , ivar(statefip) time(t_ym) gvar(pol_ym) notyet long2
csdid_estat group
returns only 2 groups

ATT by group
------------------------------------------------------------------------------
| Coefficient Std. err. z P>|z| [95% conf. interval]
-------------+----------------------------------------------------------------
GAverage | -2.229343 1.138762 -1.96 0.050 -4.461275 .0025895
G588 | 2.758401 2.317998 1.19 0.234 -1.78479 7.301593
G702 | -7.217087 1.312828 -5.50 0.000 -9.790183 -4.643991
------------------------------------------------------------------------------



Is there any one who knows how I can get the all 10 groups even though I include covariates as much as I want?

[Nathan Choi]

I just checked All ATTs, and I learned that all ATTs are calculated, but weights are calculated only for the groups that returned. Why is this happening...

[FernandoRios]

Can you post what you get with ereturn display ?
there may be a hidden bug you just happen to fins