Re: st: Endless gradient-based optimization in multilevel mixed-effects linear regression model

[Maarten buis <maartenbuis@yahoo.co.uk>]

--- On Wed, 11/8/10, Thiemann, Ulf wrote:
> I am trying to run a multilevel mixed-effects linear
> regression model using the xtmixed command. Age is one of
> the covariates among others. <snip> However, the command
> does not compute in some cases, if age is included.

It usually helps to rescale variables such that the value
zero falls within the range of the data. Remember, that the
random effects models also model the constant, that is, the 
predicted value when all covariates are zero. You can 
imagine that Stata will find it easier to produce a model 
for that when that constant represents something that is
not too far removed from your observations. 

Age is a usual suspect for such problems, as an age of 0 
means newly born babies, which are often way outside the 
range of the data. Another common problem variable is year
of birth. 

You can easily solve that by creating a new variable that 
contains age minus some reasonable constant, e.g. the age 
of the youngest observation or the average age in the 
first wave, and use that new variable instead of age.

Hope this helps,
Maarten

--------------------------
Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
Germany

http://www.maartenbuis.nl
--------------------------


      

*
*   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/