What does 0(empty) means for time dummy?

[Eve Liu]

Dear all,

I am currently running a regression with IV-FE model. My outcome is a dummy indicate whether you have the disease or not. Now, I am trying see the long-term effect of my x to the disease, I use an lagged model, my code is like:
xtivreg disease L1.age L1.age2 L1.inter1 L1.inter2 L1.marital (L1.x = L1.indicator) i.wave , first fe vce(r)

First-stage within regression

Fixed-effects (within) regression Number of obs = 25,525
Group variable: newid Number of groups = 18,365

R-squared: Obs per group:
Within = 0.1687 min = 1
Between = 0.0021 avg = 1.4
Overall = 0.0002 max = 2

F(7,18364) = 82.21
corr(u_i, Xb) = -0.4218 Prob > F = 0.0000

(Std. err. adjusted for 18,365 clusters in newid)
------------------------------------------------------------------------------
| Robust
__000004 | Coefficient std. err. t P>|t| [95% conf. interval]
-------------+----------------------------------------------------------------
L1. age | .0397042 .0104795 3.79 0.000 .0191634 .0602451

L1.age2 | .0036122 .0004886 7.39 0.000 .0026546 .0045699

L1. inter1| -.0547736 .0079861 -6.86 0.000 -.070427 -.0391202

L1.inter2 | -.0044567 .000495 -9.00 0.000 -.005427 -.0034863

L1. marital| .0146429 .0169367 0.86 0.387 -.0185547 .0478405
|
wave |
2 | 0 (empty)
5 | 0 (empty)
6 | -.0807096 .0137406 -5.87 0.000 -.1076424 -.0537768
7 | 0 (omitted)
|
L1. indicator | .2060537 .0203972 10.10 0.000 .1660734 .2460341
|
_cons | .7212392 .0451787 15.96 0.000 .6326847 .8097937
-------------+----------------------------------------------------------------
sigma_u | .53949364
sigma_e | .16508018
rho | .91438564 (fraction of variance due to u_i)
------------------------------------------------------------------------------

Fixed-effects (within) IV regression Number of obs = 73,734
Group variable: newid Number of groups = 40,602

R-squared: Obs per group:
Within = 0.0934 min = 1
Between = 0.0246 avg = 1.8
Overall = 0.0363 max = 4


Wald chi2(9) = 5429.12
corr(u_i, Xb) = 0.0183 Prob > chi2 = 0.0000

(Std. err. adjusted for 40,602 clusters in newid)
------------------------------------------------------------------------------
| Robust
heartever | Coefficient std. err. z P>|z| [95% conf. interval]
-------------+----------------------------------------------------------------
L1.x | .0095102 .0202803 0.47 0.639 -.0302385 .049259
|
L1.age | -.0022252 .0029922 -0.74 0.457 -.0080898 .0036394
|
L1.age2 | .000105 .0001101 0.95 0.340 -.0001108 .0003209

L1. inter1| .0050796 .0021004 2.42 0.016 .0009629 .0091963
|
L1.inter2 | .0001413 .0001388 1.02 0.309 -.0001308 .0004134
|
L1.marital | -.016704 .009044 -1.85 0.065 -.0344299 .0010218
|
wave |
5 | .0635051 .0143948 4.41 0.000 .0352919 .0917184
6 | .0895997 .0182082 4.92 0.000 .0539124 .1252871
7 | .1143624 .022165 5.16 0.000 .0709198 .1578051
|
_cons | .0691962 .016693 4.15 0.000 .0364785 .101914
-------------+----------------------------------------------------------------
sigma_u | .35666343
sigma_e | .16525474
rho | .8232624 (fraction of variance due to u_i)
------------------------------------------------------------------------------
Instrumented: L.x
Instruments: L.age L.age2 L.inter1 L.inter2 L.marital 5.wave
6.wave 7.wave L.indicator


My dataset contains waves from 1 to 7 except 3. As you can see in the first stage, coefficient for wave 2 and 5 are empty and I am not sure what happens here. I will appreciate if anyone can help me with it.

Best,
Eve

[Jeff Pitblado (StataCorp)]

(empty) in the coefficient table is a label that identifies factor levels that are specified in the model but are not present in the estimation sample. The different sample size for the first stage regression on L1.x of 25,525 compared to the main outcome regression 73,734 indicates to me that you have missing values in x and that the levels 2 and 5 of wave do not occur when x is not missing.

The (empty) label more commonly happens for unobserved combinations of factor variables in an interaction, such as in the following regression using the auto data where rep78 does not take on the values 1 or 2 when foreign is 1 (Foreign).

. sysuse auto
(1978 automobile data)

. regress mpg for#rep
note: 1.foreign#1b.rep78 identifies no observations in the sample.
note: 1.foreign#2.rep78 identifies no observations in the sample.

      Source |       SS           df       MS      Number of obs   =        69
-------------+----------------------------------   F(7, 61)        =      4.88
       Model |  839.550121         7  119.935732   Prob > F        =    0.0002
    Residual |  1500.65278        61  24.6008652   R-squared       =    0.3588
-------------+----------------------------------   Adj R-squared   =    0.2852
       Total |   2340.2029        68  34.4147485   Root MSE        =    4.9599

-------------------------------------------------------------------------------
          mpg | Coefficient  Std. err.      t    P>|t|     [95% conf. interval]
--------------+----------------------------------------------------------------
foreign#rep78 |
  Domestic#2  |     -1.875   3.921166    -0.48   0.634    -9.715855    5.965855
  Domestic#3  |         -2   3.634773    -0.55   0.584    -9.268178    5.268178
  Domestic#4  |  -2.555556   3.877352    -0.66   0.512     -10.3088     5.19769
  Domestic#5  |         11   4.959926     2.22   0.030     1.082015    20.91798
   Foreign#1  |          0  (empty)
   Foreign#2  |          0  (empty)
   Foreign#3  |   2.333333   4.527772     0.52   0.608    -6.720507    11.38717
   Foreign#4  |   3.888889   3.877352     1.00   0.320    -3.864357    11.64213
   Foreign#5  |   5.333333   3.877352     1.38   0.174    -2.419912    13.08658
              |
        _cons |         21   3.507197     5.99   0.000     13.98693    28.01307
-------------------------------------------------------------------------------

[Eve Liu]

Hi Jeff, thanks for the reply.

Yes, I guess this is the reason. However, in my case with the lagged model, I expect to see wave 1, 4, 5 and 6 are used for first stage estimation. So wave 2 and 7 should show with (empty) as they are in the sample but not estimated with the regression.

Also, does this influence my estimation for the key variable x here?