Statalist


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

Re: st: growth curve model with weights


From   Stas Kolenikov <skolenik@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: growth curve model with weights
Date   Thu, 18 Jun 2009 14:00:15 -0500

I don't really know if Mplus will handle the weights the way you want.
But other than that it might just compute all the
within-between-whatever matrices at different levels and run the model
very quickly with them I don't really know much more though. -gllamm-,
on the other hand, works by integrating everything numerically, and
that's why it is so slow. Sophia herslef discrecommends -gllamm- for
continuous response data. If Mplus would also have to do integration,
it would probably be through a faster binary code (rather than the
interpreted code of -gllamm-). On the other hand, on a parallel
computer, -gllamm- under Stata/MP4 or MP8 might beat Mplus since Stata
is parallelized (and I doubt Mplus is).

You should be able to get decent starting values from unweighted
analysis using -xtmixed-. Good starting values can improve estimation
time by a factor of two.

On 6/18/09, P C <calicep@yahoo.com> wrote:
> Wow, it does take a lot of time. I heard about the negative side of gllamm. I am wondering whether I should try Mplus. Do you know if it does a better job (time-wise) for this kind of modelling?
>
>  There is a need for special handling of the school-level weights. I am trying to understand it from her paper but it's not just as simple as scaling.

-- 
Stas Kolenikov, also found at http://stas.kolenikov.name
Small print: I use this email account for mailing lists only.
*
*   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