Re: st: Extremely High Correlation between Interaction Term and Each of Its Component Linear Terms

[Richard Williams <richardwilliams.ndu@gmail.com>]

At 04:01 AM 12/12/2012, Johan Hellström wrote:
> Lloyd,
>
> Yes, you can save yourself from the trouble of 
> worrying about collinearity caused by 
> interacting terms. Interaction terms and its 
> components are usually correlated and are so 
> correctly. There is no magical trick to reduce 
> the collinearity and although centring variables 
> could be helpful sometimes to ease 
> interpretation, it does not alter anything statistically or substantively.
>
> Best wishes,
> Johan Hellstrom

Yes, I don't worry too much about 
multicollinearity in such instances, unless maybe 
the model won't converge. For example, if you 
have x and x^2 in the model, you might need to 
rescale x if x^2 produces monstrous numbers. So, 
for example, it may be better to measure income 
in thousands of dollars rather than in dollars.

The handout that Lloyd mentioned earlier,

http://www.nd.edu/~rwilliam/stats2/l11.pdf

does include some tips for dealing with 
multicollinarity. For example, avoid doing 
something stupid like including a scale that was 
computed from other variables in your model. Or, 
consider creating a scale if, say, x1-x10 are 
really just 10 different ways of measuring the same concept.

Other than that, always collecting samples of at 
least 10 million cases is probably a good idea. 
Assuming you have unlimited resources for your research.


-------------------------------------------
Richard Williams, Notre Dame Dept of Sociology
OFFICE: (574)631-6668, (574)631-6463
HOME:   (574)289-5227
EMAIL:  Richard.A.Williams.5@ND.Edu
WWW:    http://www.nd.edu/~rwilliam


*
*   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/