# st: GLLAMM - flat or discontinuous region encountered (binary, logit)

 From [email protected] To [email protected] Subject st: GLLAMM - flat or discontinuous region encountered (binary, logit) Date Wed, 10 Jan 2007 16:49:43 -0200

```Hello.

I am working on a case of using GLLAMM for structural equation modeling with a
binary outcome - it is a path model, with one latent variable, and I received
the message (Stata 9.2):

"
numerical derivatives are approximate
flat or discontinuous region encountered
"

===============================================================================

. gllamm fetsobrev sevar, i(nordem) family(binomial) link(logit) nocons

General model information
-----------------------------------------------------------------------------
dependent variable:         fetsobrev
family:                     binom
denominator:                1
equation for fixed effects  sevar

Random effects information for 2 level model
-----------------------------------------------------------------------------

***level 2 (nordem) equation(s):
(1 random effect(s))

standard deviation for random effect 1
nord1 : sevar

number of level 1 units = 3339
number of level 2 units = 477

Initial values for fixed effects

Iteration 0:   log likelihood = -2314.4184
Iteration 1:   log likelihood = -2268.9904
Iteration 2:   log likelihood = -2268.9636

Logistic regression                               Number of obs   =       3339
LR chi2(1)      =          .
Log likelihood = -2268.9636                       Prob > chi2     =          .

------------------------------------------------------------------------------
fetsobrev |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
sevar |  -.5529417   .0590998    -9.36   0.000    -.6687751   -.4371083
------------------------------------------------------------------------------

start running on 10 Jan 2007 at 18:42:45

Updating posterior means and variances
log-likelihood:
-2208.5065
------------------------------------------------------------------------------
Iteration 0:
Coefficient vector:
fetsobrev:      nord1:
sevar       sevar
r1   -.5529417          .5

log likelihood = -2208.5065
(not concave)
------------------------------------------------------------------------------
Iteration 1:
Coefficient vector:
fetsobrev:      nord1:
sevar       sevar
r1   -3.468325    29.05551

log likelihood = -1773.4122
------------------------------------------------------------------------------
Iteration 2:
Coefficient vector:
fetsobrev:      nord1:
sevar       sevar
r1   -25.39545    -74.3894

log likelihood =  -1754.878
------------------------------------------------------------------------------
Iteration 3:
Coefficient vector:
fetsobrev:      nord1:
sevar       sevar
r1    -55.6269   -157.9688

log likelihood = -1753.0193
------------------------------------------------------------------------------
Iteration 4:
Coefficient vector:
fetsobrev:      nord1:
sevar       sevar
r1   -57.03351   -161.8858

log likelihood = -1753.0178
------------------------------------------------------------------------------
Iteration 5:
Coefficient vector:
fetsobrev:      nord1:
sevar       sevar
r1    -57.7367   -163.8439

log likelihood = -1753.0171
------------------------------------------------------------------------------
Iteration 6:
Coefficient vector:
fetsobrev:      nord1:
sevar       sevar
r1   -58.08837   -164.8232

log likelihood = -1753.0168
------------------------------------------------------------------------------
Iteration 7:
Coefficient vector:
fetsobrev:      nord1:
sevar       sevar
r1   -58.26419   -165.3129

log likelihood = -1753.0166
------------------------------------------------------------------------------
Iteration 8:
Coefficient vector:
fetsobrev:      nord1:
sevar       sevar
r1   -58.35211   -165.5577

log likelihood = -1753.0165
(backed up)
------------------------------------------------------------------------------
Iteration 9:
Coefficient vector:
fetsobrev:      nord1:
sevar       sevar
r1   -58.39607   -165.6801

log likelihood = -1753.0165
(backed up)
------------------------------------------------------------------------------
Iteration 10:
Coefficient vector:
fetsobrev:      nord1:
sevar       sevar
r1   -58.41804   -165.7413

log likelihood = -1753.0165
(backed up)
------------------------------------------------------------------------------
Iteration 11:
Coefficient vector:
fetsobrev:      nord1:
sevar       sevar
r1   -58.42355   -165.7566

numerical derivatives are approximate
flat or discontinuous region encountered
log likelihood = -1753.0165
(backed up)
------------------------------------------------------------------------------
Iteration 12:
Coefficient vector:
fetsobrev:      nord1:
sevar       sevar
r1   -58.47534    -165.756

numerical derivatives are approximate
flat or discontinuous region encountered
log likelihood = -1753.0163
------------------------------------------------------------------------------

(and so on until get always the same coefficients and no change in likelihood
...)

Iteration 19:
Coefficient vector:
fetsobrev:      nord1:
sevar       sevar
r1   -58.48807   -165.7559

numerical derivatives are approximate
flat or discontinuous region encountered
numerical derivatives are approximate
flat or discontinuous region encountered
log likelihood = -1753.0162
(backed up)

===============================================================================

I found an old message where someone with the same problem but there was a
single observation for each client and this is not my case.

Any help apreciated.

Gizelton.
FSP / USP

-----

The message I read before is available in

<http://www.stata.com/statalist/archive/2002-09/msg00456.html> :

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
```