st: Multicollinearity/VIF using -xtlogit,re- models

[Andreas Schiffelholz <Andreas.Schiffelholz@gmx.de>]

Hello everyone,

I'm working with an unbalanced panel dataset (t: 10 years, x: 170 
companies) calculating -xtlogit- regressions with random effects 
including "normal" (company age, sales growth, ...) as well as dummy 
variables (fiscal year, industry). Now I want to check these models for 
multicollinearity.  I want to calculate the VIFs using the -collin- 
command. Is there anything I need to consider because I'm working with a 
panel model or is there no difference between the VIFs of a standard 
-logit- and an -xtlogit, re- random effects model?

There is one specific source of multicollinearity I'm curious about: I'm 
using a company age (in years) variable as well as dummies for the 
different fiscal years. The VIFs calculated for "company age" do not 
indicate any problems with multicollinearity and I receive significant 
results for the "company age" variable in my -xtlogit,re- model. On the 
other hand as far as I understand the within variance of the "company 
age" variables should be explained perfectly by the fiscal year dummies. 
Do I have to change the "company age" variable and what would be the 
best way to do so? Is there a specific standard what kind of variable to 
use? (e.g. variable "company age in 2002" or "company age when entering 
the sample")

thanks a lot in advance.
Andreas Schiffelholz



*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/