Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: gllamm convergence problem


From   Marco Alfano <m.u.alfano@googlemail.com>
To   statalist@hsphsun2.harvard.edu
Subject   st: gllamm convergence problem
Date   Sun, 28 Jun 2009 18:58:35 +0100

Dear Statalisters,



I hope you all are doing well. I am having some problems to get my
model to converge. It does converge for some subsamples but generally
the models do not converge.



I am estimating a covariate measurement error model using gllamm. The
measurements of the latent factor are dichotomous variables (so the
measurement model is a two parameter item response model). The outcome
model is a discrete time duration model, which I am estimating by
expanding the data according to the duration variable. I am
particularly interested in the factor loading on the latent variable
z3.



Once the dataset is reshaped and expanded the commands for the
estimations look like the following:



***************************************************************************

***************************************************************************



ge z_x1 = z3*x1

ge z_x2 = z3*x2

ge z_x3 = z3*x3

ge z_x4 = z3*x4



*** Latent Equation and True Covariate model



eq latentfactors: z1 z2 z3



eq truecovariatemodel: x1 x2 x3 x4 x5



*** Outcome model



gllamm y z1 z2 z3 z_x1 z_x2 z_x3 z_x4 , ///

i(kids_id) eqs(latentfactors) link(logit logit) family(binom binom) ///

lv(measuredis) fv(measuredis) geqs(truecovariatemodel)



***************************************************************************

***************************************************************************



I would really appreciate some advice or general tips about convergence.



Thanks a lot already to everyone and have a good Sunday



marco.
*
*   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/



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