:
. list id trial trial10 response
+----------------------------------+
| id trial trial10 response |
|----------------------------------|
1. | 1 1 -1 0 |
2. | 1 2 -1 0 |
3. | 1 3 -1 1 |
4. | 1 4 -1 1 |
5. | 1 5 -1 0 |
|----------------------------------|
6. | 1 6 -1 1 |
7. | 1 7 -1 1 |
8. | 1 8 -1 1 |
9. | 1 9 -1 0 |
10. | 1 10 9 1 |
|----------------------------------|
11. | 1 11 0 0 |
12. | 1 12 0 0 |
13. | 1 13 0 1 |
14. | 2 1 -1 1 |
15. | 2 2 -1 0 |
|----------------------------------|
16. | 2 3 -1 1 |
17. | 2 4 -1 0 |
18. | 2 5 -1 0 |
19. | 2 6 -1 1 |
20. | 2 7 -1 0 |
|----------------------------------|
21. | 2 8 -1 1 |
22. | 2 9 -1 0 |
23. | 2 10 9 1 |
24. | 2 11 0 0 |
25. | 2 12 0 1 |
|----------------------------------|
26. | 2 13 0 0 |
:
. xtmelogit response trial10 if condi == 1 || id:, or
Refining starting values:
Iteration 0: log likelihood = -2029.4266 (not concave)
Iteration 1: log likelihood = -1948.395
Iteration 2: log likelihood = -1941.9592
Performing gradient-based optimization:
Iteration 0: log likelihood = -1941.9592
Iteration 1: log likelihood = -1941.0187
Iteration 2: log likelihood = -1941.0171
Iteration 3: log likelihood = -1941.0171
Mixed-effects logistic regression Number of obs = 2821
Group variable: id Number of groups = 217
Obs per group: min = 13
avg = 13.0
max = 13
Integration points = 7 Wald chi2(1) = 2.63
Log likelihood = -1941.0171 Prob > chi2 = 0.1048
------------------------------------------------------------------------------
response | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
trial10 | .9770401 .0139913 -1.62 0.105 .9499989 1.004851
_cons | 1.212697 .0458986 5.10 0.000 1.125993 1.306078
------------------------------------------------------------------------------
------------------------------------------------------------------------------
Random-effects Parameters | Estimate Std. Err. [95% Conf. Interval]
-----------------------------+------------------------------------------------
id: Identity |
sd(_cons) | 6.50e-12 .0748592 0 .
------------------------------------------------------------------------------
LR test vs. logistic regression: chibar2(01) = 3.0e-11 Prob>=chibar2 = 1.0000
What confuses me is the estimate for sd(_cons) under random-effects parameters. Why does Stata not compute a 95% CI? I also find it strange that the test statistic for the LR test vs. logistic regression shown at the very bottom is almost 0 with a p-value of 1.0000. Do these results mean that the model is not converging, or do they say something meaningful about my data? It seems implausible to me that every participant would have the same intercept.
:
. xtmelogit response trial10 if condi == 1 || id:trial10, or
Refining starting values:
Iteration 0: log likelihood = -2090.5208 (not concave)
Iteration 1: log likelihood = -1954.3232
Iteration 2: log likelihood = -1944.9755
Performing gradient-based optimization:
Iteration 0: log likelihood = -1944.9755
Iteration 1: log likelihood = -1941.0213
Iteration 2: log likelihood = -1941.0171
Iteration 3: log likelihood = -1941.0171
Mixed-effects logistic regression Number of obs = 2821
Group variable: id Number of groups = 217
Obs per group: min = 13
avg = 13.0
max = 13
Integration points = 7 Wald chi2(1) = 2.63
Log likelihood = -1941.0171 Prob > chi2 = 0.1048
------------------------------------------------------------------------------
response | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
trial10 | .9770401 .0139913 -1.62 0.105 .9499989 1.004851
_cons | 1.212697 .0458986 5.10 0.000 1.125993 1.306078
------------------------------------------------------------------------------
------------------------------------------------------------------------------
Random-effects Parameters | Estimate Std. Err. [95% Conf. Interval]
-----------------------------+------------------------------------------------
id: Independent |
sd(trial10) | 6.31e-12 .0553508 0 .
sd(_cons) | 1.71e-11 .0748591 0 .
------------------------------------------------------------------------------
LR test vs. logistic regression: chi2(2) = 3.0e-11 Prob > chi2 = 1.0000
I'd appreciate any insight you might have about what this means -- and if it's an estimation problem, how to fix it. Thanks very much in advance for your time.