Statalist


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

Re: st: GLLAMM modeling


From   "Stas Kolenikov" <skolenik@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: GLLAMM modeling
Date   Mon, 20 Aug 2007 21:36:55 -0500

1e8 is probably large. 1e3 is unpleasant, but within the "typical"
range. It affects (i) the efficiency of nonlinear iteration steps, and
when extremely large, (ii) numerical accuracy. Very high number may be
a symptom of underidentification, although masked by nonlinearities
and approximations used by -gllamm-. To reduce this condition number,
try orthogonalizing and standardizing everything, so that variables do
not have wildly different scales and are centered nicely, although
that may help only partially.

On 8/20/07, BORHANI, HOSSEIN <hborhani@ersgroup.com> wrote:
> In GLLAMM manual regarding the "condition number" that appears in
> stata's output after you run a GLLAMM model it states:
>
> "The condition number will be large when the Hessian matrix is nearly
> singular indicating that the model may not well identified"
>
> What constitutes a "large" or a "low" condition number? Is there any
> criteria?
>
> Thank you,
> Borhani
>
> *
> *   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/
>


-- 
Stas Kolenikov, also found at http://stas.kolenikov.name

Small print: Please do not reply to my Gmail address as I don't check
it regularly.
*
*   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