meologit - understanding when / if var(_cons) is a problem

[Nora Romeo]

Hello,

I have a mixed-effects model example I am supposed to interpret. In the example, the random effects variance (var(_cons)) is many times higher than any of the coefficients for the fixed effects. Does this suggest a fundamental problem in the model, or how should I interpret this?

Thanks!

Fitting fixed-effects model:

Iteration 0: log likelihood = -144.6264
Iteration 1: log likelihood = -144.06933
Iteration 2: log likelihood = -144.06848
Iteration 3: log likelihood = -144.06848

Refining starting values:

Grid node 0: log likelihood = -139.35779

Fitting full model:

Iteration 0: log pseudolikelihood = -139.35779
Iteration 1: log pseudolikelihood = -137.23535
Iteration 2: log pseudolikelihood = -137.01036
Iteration 3: log pseudolikelihood = -137.00532
Iteration 4: log pseudolikelihood = -137.00529
Iteration 5: log pseudolikelihood = -137.00529

Mixed-effects logistic regression Number of obs = 269
Group variable: C0_code Number of groups = 213

Obs per group:
min = 1
avg = 1.3
max = 2

Integration method: mvaghermite Integration pts. = 7

Wald chi2(6) = 19.66
Log pseudolikelihood = -137.00529 Prob > chi2 = 0.0032
(Std. Err. adjusted for 213 clusters in C0_code)
------------------------------------------------------------------------------
| Robust
move_n | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
events|
1 | 1.20317 .6106109 0.36 0.716 .44498 3.253219
2 | 5.434769 7.008151 1.31 0.189 .4340643 68.04686
3 | 8.965599 9.314639 2.11 0.035 1.170145 68.69402
|
controla| 4.184571 3.344236 1.79 0.073 .8737466 20.04086
2.C0_time | .1017764 .0787512 -2.95 0.003 .0223366 .463744
control2| 2.439937 .7012192 3.10 0.002 1.389152 4.285559
_cons | .0138684 .0171629 -3.46 0.001 .0012264 .1568322
-------------+----------------------------------------------------------------
C0_code |
var(_cons)| 63.478622 1.685231 1.345983 8.990316
------------------------------------------------------------------------------

[Carlo Lazzaro]

Nora:
are you sure that the folllowing data:

 
 var(_cons)| 63.478622 1.685231 1.345983 8.990316

are reported exactly as they appear in -meologit- outcome table?

[Clyde Schechter]

Carlo Lazzaro is right to be suspicious. The estimate itself, 63.478... should fall within the confidence interval (1.345..., 8.990...) which it clearly does not. Something is amiss.