Statalist The Stata Listserver


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: =?UTF-8?Q?problem=20with=20log-likelihood=20in=20GLLAMM?=


From   =?UTF-8?Q?Janek=20Nowak?= <janek_nowak@o2.pl>
To   statalist@hsphsun2.harvard.edu
Subject   st: =?UTF-8?Q?problem=20with=20log-likelihood=20in=20GLLAMM?=
Date   Wed, 06 Sep 2006 21:26:19 +0200

Dear Stata users,

I am trying to estimate a Heckman sample selection model for panel data in Stata using GLLAMM.
I encountered the following problem.

Sometimes the log likelihood instead of increasing decrease. I mean that for some 'i's iteration[i] log likelihood > iteration[i+1] log likelihood (as you can see in attached printscreen for iteration[5] and iteration [6]). 
Does it mean that I should try with more quadrature points or there is a problem with the data? Is it possible that despite this 'jumping' the model can still converge or I should rather abort maximalization process.

Running adaptive quadrature
Iteration 0:    log likelihood = -8724.0818
Iteration 1:    log likelihood = -7729.0175
Iteration 2:    log likelihood = -6772.7395
Iteration 3:    log likelihood = -6420.9296
Iteration 4:    log likelihood = -6410.1388
Iteration 5:    log likelihood = -6111.5488
Iteration 6:    log likelihood =  -6129.489
Iteration 7:    log likelihood = -6122.5803
Iteration 8:    log likelihood = -6118.6587
Iteration 9:    log likelihood = -6113.8941
Iteration 10:   log likelihood = -6127.7954

I would really appreciate any advice.

Thanks,
Janek


*
*   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