Statalist The Stata Listserver


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

Re: st: nested design and logistic regression


From   Joseph Coveney <jcoveney@bigplanet.com>
To   Statalist <statalist@hsphsun2.harvard.edu>
Subject   Re: st: nested design and logistic regression
Date   Fri, 07 Jul 2006 10:31:26 +0900

Denis Haine wrote:

Thanks Joseph for your advice. I have tried gllamm before asking the list
but was not sure how to setup the model and was a bit reluctant to use this
difficult feature of Stata (for me!). The gllamm model worked with my data
but, as a newby to gllamm, I don't really understand your last lines: the
purpose of setting up a matrix and re-run the gllamm command.

--------------------------------------------------------------------------------

The GLLAMM Manual ( www.bepress.com/ucbbiostat/paper160/ ) illustrates far
more important uses of -gllamm-'s -from()- option, but in my case, the
purpose was to save time.  -gllamm- uses eight integration points by
default, which runs faster than with 30.  If there is a problem with
convergence, then it will be detected sooner with a run with fewer
integration points, so I usually do the first run with eight (often fewer)
integration points.  If things go well, I retain the fit in the matrix for
use as initial parameter estimates for the run with a higher number of
integration points, usually accompanied by -quadchk- like examinations.  The
_GLLAMM Manual_ shows the use of the -from()- option in the context of
assessing the adequacy of the number of quadrature points on pp. 37--38 and
pp. 81--82.

It dawned on me that your type of analysis requirement might be covered in
S. Rabe-Hesketh & A. Skrondal, _Multilevel and Longitudinal Modeling using
Stata_ (College Station, Texas: Stata Press, 2005).  I have to confess that
I actually have bought the book, but didn't think to check it before
responding to your post.  I don't have it at-hand at the moment, but I'd bet
dollars to doughnuts that it has something pertinent.  Well worth the
purchase price in any event, especially to get an overview of what's
available in Stata to fit mixed models and to get insight on how best to
work the commands.  Note that there have been new user-written commands
since publication (e.g., -regoprob-), and I'm hoping that some
of -xtmixed-'s to-do list's items* on Slide 17 of Bobby Gutierrez's
presentation last year at the U.K. Stata Users' Group meeting will have been
crossed-off by the time of his workshop this September in Melbourne
following the Australian and New Zealand Stata Users' Group (ANZSUG)
meeting.  Bobby's workshop will undoubtedly be well worth attending,
regardless.

Joseph Coveney

*Dear Santa:  please add -constraints()-.  Thank you.

*
*   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