Hi,
I am trying to use reoprob on several subsets of data, but am frequently
returning an error that says "no observations," when there are in fact a
few hundred or thousand observations.
Here is a sample of what has worked and has not:
. reoprob healthstatus_ mothed_1997 female age_1997 black hispanic
otherrace imm bir
> thwt lnincome_ lcstate_1997 after97 lcstate_after97 if
insuredcurrent_1997==0, i(
> id)
Fitting constant-only model:
Iteration 0: log likelihood = -210.53583
Iteration 1: log likelihood = -203.35242
Iteration 2: log likelihood = -202.89348
Iteration 3: log likelihood = -202.88512
Iteration 4: log likelihood = -202.88512
Fitting full model:
Iteration 0: log likelihood = -193.35161
Iteration 1: log likelihood = -185.89095
Iteration 2: log likelihood = -185.42578
Iteration 3: log likelihood = -185.41409
Iteration 4: log likelihood = -185.41409
Random Effects Ordered Probit Number of obs
= 166
LR chi2(12) =
34.94
Log likelihood = -185.41409 Prob > chi2 =
0.0005
------------------------------------------------------------------------------
healthstat~_ | Coef. Std. Err. z P>|z| [95% Conf.
Interval]
-------------+----------------------------------------------------------------
eq1 |
mothed_1997 | -.1343246 .078087 -1.72 0.085 -.2873724
.0187232
female | -.1134465 .3173086 -0.36 0.721 -.7353599
.508467
age_1997 | .0762218 .0500739 1.52 0.128 -.0219213
.1743649
black | .9444976 .4461865 2.12 0.034 .0699881
1.819007
hispanic | -1.915888 1.679852 -1.14 0.254 -5.208337
1.37656
otherrace | -11.4049 8806.956 -0.00 0.999 -17272.72
17249.91
imm | 1.552197 1.595893 0.97 0.331 -1.575695
4.68009
birthwt | .0041629 .0327088 0.13 0.899 -.0599452
.068271
lnincome_ | .0494158 .1784086 0.28 0.782 -.3002586
.3990902
lcstate_1997 | -.1160115 .3668058 -0.32 0.752 -.8349378
.6029147
after97 | .0468742 .3186625 0.15 0.883 -.5776929
.6714413
lcstate_a~97 | -1.064486 .4077225 -2.61 0.009 -1.863607
-.2653644
-------------+----------------------------------------------------------------
_cut1 |
_cons | -.5617763 1.932602 -0.29 0.771 -4.349607
3.226054
-------------+----------------------------------------------------------------
_cut2 |
_cons | .7659718 1.940549 0.39 0.693 -3.037434
4.569377
-------------+----------------------------------------------------------------
_cut3 |
_cons | 2.267353 1.967665 1.15 0.249 -1.589199
6.123905
-------------+----------------------------------------------------------------
_cut4 |
_cons | 3.164919 2.00701 1.58 0.115 -.7687487
7.098586
-------------+----------------------------------------------------------------
rho |
_cons | .503459 .1131211 4.45 0.000 .2817458
.7251722
------------------------------------------------------------------------------
reoprob healthstatus_ mothed_1997 female age_1997 black hispanic
otherrace imm bir
> thwt lnincome_ lcstate_1997 after97 lcstate_after97 if
insuredcurrent_1997==1, i(
> id)
no observations
r(2000);
The second statement is exactly the same as the first, except for the
"if" statement. The sub-sample in the first (working) regression has
302 ids, while the sub-sample in the second (non-working) regression has
2500. Any idea why this is happening?
Thanks,
Caroline
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/