Re: Antwort: Re: Antwort: Re: st: Multicollinearity in panel data

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

--- On Wed, 9/2/11, Justina Fischer wrote:
> From my own experience, it works for age, age squared and age
> cubed (all not demeaned), if some people are sufficiently
> 'aged' in the data

I like centering my variables, not necessarily at the mean, but
at least such that 0 falls within or near the observed range of
my variables. That way the constant would be the expected value
of the outcome variable for some typical individual as defined 
by the way I centered my variables, and I usually have a pretty 
good idea of what that should be. This saved me couple of times 
from some pretty nasty coding mistakes... Also this gives a 
natural way to introduce in the text a baseline against which 
one can compare the size of the effects. This can help when 
discussing what effects are "big" or "small". Admittedly, this 
is more a literary trick than a statistical one, but every 
little bit helps.

-- Maarten

Ps. I would not center age^2 or age^3, I would center age, and
square and cube the centered age variable.

--------------------------
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/