Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down at the end of May, and its replacement, statalist.org is already up and running.


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

Re: st: Estimation metric of variance components in gllamm


From   Stas Kolenikov <skolenik@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Estimation metric of variance components in gllamm
Date   Sat, 23 Jun 2012 09:04:09 -0500

-gllamm- uses Cholesky decomposition. So I think your best bet is to
form the covariance matrix of the random effects, element-by-element,
Cholesky-decompose it, and feed that to -gllamm-. I thought that
-gllamm- was using the log of variances, too, BTW.

In my experience with -gllamm-, it is the convergence of the random
effect parameters that takes the longest. So if you can provide really
good starting values, you should be able to cut the iteration time.

On Sat, Jun 23, 2012 at 3:41 AM, Sascha Peter
<sascha.peter@uni-hamburg.de> wrote:
> Dear Stata-Users,
> I have a question regarding the estimation metric of random effects
> covariances in gllamm.
> I want to use estimation results from Stata's xtmepoisson as starting values
> to estimate a multilevel Poisson regression in gllamm. Since the estimation
> metric for the variance components of the two programs differ, variances and
> covariances stored in e(b) after xtmepoisson have to be transformed before
> these can be used as starting values for gllamm.
> xtmepoisson uses the log of the standard deviation as estimation metric and
> gllamm uses the standard deviation. So exponentiation of the respective
> elements in e(b) will do the job.
> xtmepoisson uses the arc-hyperbolic tangents of correlations to estimate
> covariance components. However,I was not able to figure out in which metric
> these are estimated in gllamm. Which transformation has to be applied to the
> covariance components in e(b) so that these estimates can be used as
> starting values in gllamm?
>
> Thank you very much,
> Sascha
>
> *
> *   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/



-- 
---- Stas Kolenikov
-- http://stas.kolenikov.name
---- Senior Survey Statistician, Abt SRBI
-- Opinions stated in this email are mine only, and do not reflect the
position of my employer

*
*   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   |   Site index