Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: GLLAMM - a problem between Stata 8 and Stata 9


From   Jean-Benoit Hardouin <jean-benoit.hardouin@orscentre.org>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: GLLAMM - a problem between Stata 8 and Stata 9
Date   Fri, 17 Jun 2005 23:55:11 +0200

The same problem exists with a vector of estimations obtained with GLLAMM :

/******ESTIMATION OF THE PARAMETERS WITH GLLAMM*****************/

. gllamm rep rep1-rep9, i(id) nocons family(binom) link(logit) trace

General model information
------------------------------------------------------------------------------

dependent variable: rep
family: binom
link: logit
denominator: 1
equation for fixed effects rep1 rep2 rep3 rep4 rep5 rep6 rep7 rep8 rep9

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


***level 2 (id) equation(s):

standard deviation of random effect
id1: _cons

number of level 1 units = 4248
number of level 2 units = 472

...

number of level 1 units = 4248
number of level 2 units = 472

Condition Number = 3.4788365

gllamm model

log likelihood = -2207.4371

------------------------------------------------------------------------------
rep | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
rep1 | -.1174269 .1311706 -0.90 0.371 -.3745166 .1396628
rep2 | 2.496076 .172519 14.47 0.000 2.157945 2.834207
rep3 | 2.38552 .1686237 14.15 0.000 2.055024 2.716016
rep4 | 1.081889 .1382254 7.83 0.000 .8109724 1.352806
rep5 | .3302297 .1318433 2.50 0.012 .0718216 .5886378
rep6 | 1.164174 .139379 8.35 0.000 .8909964 1.437352
rep7 | .8692891 .1356834 6.41 0.000 .6033545 1.135224
rep8 | 1.938048 .1551466 12.49 0.000 1.633966 2.24213
rep9 | -.1657712 .1312288 -1.26 0.207 -.422975 .0914326
------------------------------------------------------------------------------


Variances and covariances of random effects
------------------------------------------------------------------------------


***level 2 (id)

var(1): 2.2496163 (.23751741)
------------------------------------------------------------------------------

/**********************I SAVE THE ESTIMATED PARAMETERS **************************/

. matrix est2=e(b)

/********************I EVALUATE THE MODEL WITH THE ESTIMATED PARAMETERS ************/

. gllamm rep rep1-rep9, i(id) nocons family(binom) link(logit) trace from(est2) eval

General model information
------------------------------------------------------------------------------

dependent variable: rep
family: binom
link: logit
denominator: 1
equation for fixed effects rep1 rep2 rep3 rep4 rep5 rep6 rep7 rep8 rep9

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



***level 2 (id) equation(s):

standard deviation of random effect
id1: _cons

number of level 1 units = 4248
number of level 2 units = 472
------------------------------------------------------------------------------


start running on 17 Jun 2005 at 23:51:40
log-likelihood = -2207.4371
finish running on 17 Jun 2005 at 23:51:40


number of level 1 units = 4248
number of level 2 units = 472

Condition Number could not be computed

gllamm model

log likelihood = -2207.4371

last estimates not found
r(301);

/***********************************END**********************************/


Jean-Benoit Hardouin a écrit :


Dear all,
I have a dataset of binary variables (response to 9 binary items). I explain the responses to these items by a logistic mixed model. I have a program who estimates the parameters of the model with xtlogit, and evaluates the model under GLLAMM (with options from(matrix) and eval). [my aim is to obtain estimations of the random variable for each individual after GLLAMM with GLLAPRED and to gain time by using xtlogit instead of GLLAMM]
A simpler version of this program is given below.

My problem is that there is no problem to evaluate the model with GLLAMM under Stata 8 but I have an error under Stata 9 for exactly the same program.
The return error is "last estimates not found (r(301)" and I do not understand what is the problem.

Jean-Benoit Hardouin

The program is :

/*****************************BEGINING OF THE PROGRAM*******************************************/

use "http://anaqol.free.fr/statalist/jbh.dta";
matrix est=(-.1059915,2.511868,2.4014077,1.0965434,.34265641,1.1790977,.88322999,1.9542748,-.15445884,1.5169531)
local colnames
forvalues i=1/9 {
local colnames "`colnames' rep:rep`i'"
}
local colnames "`colnames' id1:_cons"
di "`colnames'"
matrix colnames est=`colnames'

matrix list est
gllamm rep rep1-rep9, i(id) nocons family(binom) link(logit) trace from(est) eval

/*****************************END OF THE PROGRAM*******************************************/


/***************************RESULTS UNDER STATA 8 ******************************************/
General model information
------------------------------------------------------------------------------

dependent variable: rep
family: binom
link: logit
denominator: 1
equation for fixed effects rep1 rep2 rep3 rep4 rep5 rep6 rep7 rep8 rep9

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



***level 2 (id) equation(s):

standard deviation of random effect
id1: _cons

number of level 1 units = 4248
number of level 2 units = 472
------------------------------------------------------------------------------


start running on 17 Jun 2005 at 23:23:35
log-likelihood = -2207.4695
finish running on 17 Jun 2005 at 23:23:36


number of level 1 units = 4248
number of level 2 units = 472

Condition Number could not be computed

gllamm model

log likelihood = -2207.4695

------------------------------------------------------------------------------
rep | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
rep1 | -.1059915 . . . . .
rep2 | 2.511868 . . . . .
rep3 | 2.401408 . . . . .
rep4 | 1.096543 . . . . .
rep5 | .3426564 . . . . .
rep6 | 1.179098 . . . . .
rep7 | .88323 . . . . .
rep8 | 1.954275 . . . . .
rep9 | -.1544588 . . . . .
------------------------------------------------------------------------------


Variances and covariances of random effects
------------------------------------------------------------------------------


***level 2 (id)

var(1): 2.3011467 (0)
------------------------------------------------------------------------------

/***************************RESULTS UNDER STATA 9 ******************************************/
General model information
------------------------------------------------------------------------------

dependent variable: rep
family: binom
link: logit
denominator: 1
equation for fixed effects rep1 rep2 rep3 rep4 rep5 rep6 rep7 rep8 rep9

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



***level 2 (id) equation(s):

standard deviation of random effect
id1: _cons

number of level 1 units = 4248
number of level 2 units = 472
------------------------------------------------------------------------------


start running on 17 Jun 2005 at 23:29:04
log-likelihood = -2207.4695
finish running on 17 Jun 2005 at 23:29:05


number of level 1 units = 4248
number of level 2 units = 472

Condition Number could not be computed

gllamm model

log likelihood = -2207.4695

last estimates not found
r(301);

end of do-file
r(301);


--
***************************************************************************
Jean-Benoit Hardouin
Biostatisticien
Observatoire Régional de la Santé du Centre
BP 2439
1, rue Porte Madeleine
45032 Orléans Cedex 1
tél : 02 38 74 48 80
fax : 02 38 74 48 81
Email : jean-benoit.hardouin@orscentre.org
**************************************************************************

*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index